Machine learning reveals prominent spontaneous behavioral changes and treatment efficacy in humanized and transgenic Alzheimer's disease models
Open accessAbstract
Computer-vision and machine-learning (ML) approaches are being developed to provide scalable, unbiased, and sensitive methods to assess mouse behavior. Here, we used the ML-based variational animal motion embedding (VAME) segmentation platform to assess spontaneous behavior in humanized App knockin and transgenic APP models of Alzheimer's disease (AD) and to test the role of AD-related neuroinflammation in these behavioral manifestations. We found marked alterations in spontaneous behavior in App NL-G-F and 5xFAD mice, including age-dependent changes in motif utilization, disorganized behavioral sequences, increased transitions, and randomness. Notably, blocking fibrinogen-microglia interactions in 5xFAD- Fgg γ390–396A mice largely prevented spontaneous behavioral alterations, indicating a key role for neuroinflammation. Thus, AD-related spontaneous behavioral alterations are prominent in knockin and transgenic models and sensitive to therapeutic interventions. VAME outcomes had higher specificity and sensitivity than conventional behavioral outcomes. We conclude that spontaneous behavior effectively captures age- and sex-dependent disease manifestations and treatment efficacy in AD models.
AI snapshot
Written from the paper’s full text.
TL;DR
Using the ML-based VAME platform, this study found prominent spontaneous behavioral alterations in humanized AppNL-G-F knockin and 5xFAD transgenic Alzheimer’s disease mouse models, with higher sensitivity and specificity than conventional open-field measures. Blocking fibrinogen-microglia interactions in 5xFAD-Fggγ390–396A mice largely prevented these behavioral alterations, implicating neuroinflammation.
Key findings
- VAME identified 30 postural motifs; 13-month-old AppNL-G-F mice showed significant alterations in 8 of 30 motifs (q < 0.05), while 6-month-old AppNL-G-F mice did not display deficits.
- AppNL-G-F mice exhibited disorganized behavioral sequences, increased transitions, and increased randomness; behavioral sequences were less predictable at Markov orders 0–3, particularly in females.
- 5xFAD mice had prominent ML behavioral alterations in 17 of 30 motifs, and nearly all were ameliorated in 5xFAD-Fggγ390–396A mice; no sex effects were seen in 5xFAD and 5xFAD-Fggγ390–396A mice.
- VAME outcomes classified AppNL-G-F vs WT with 92.9% specificity and 83.3% sensitivity, versus conventional open-field outcomes at 50.0% specificity and 77.8% sensitivity; for the 5xFAD cohort, VAME categorized 100% of mice correctly.
Methods
- Spontaneous behavior was recorded in an open arena from below at 25 frames per second for 25 min (AppNL-G-F cohorts) or 60 min (5xFAD-Fggγ390–396A cohort); DeepLabCut tracked nine body parts, and VAME segmented behavior into 30 motifs using egocentric coordinates of seven body parts.
- Cohorts included 13-month-old AppNL-G-F mice (n = 18) and WT littermates (n = 14), 6-month-old AppNL-G-F mice (n = 13) and WT controls (n = 11), 22-month-old AppNL-G-F mice (n = 17) and WT controls (n = 22) for Morris water maze, and 8–10-month-old 5xFAD (n = 13), 5xFAD-Fggγ390–396A (n = 12), Fggγ390–396A (n = 9), and WT (n = 13) mice.
- Main analyses included FDR-BH correction, two-way ANOVA, repeated-measures ANOVA, Markov chain models, transition network analysis, delta indices for transitions/use/speed, and logistic regression classifiers comparing VAME and conventional open-field outcomes.
Limitations
- The paper states that specific behavioral domains (e.g., cognitive functions) and functional brain topology were not explicitly assessed, and future studies are needed to determine contributions of distinct behavioral domains, neural systems, or brain regions.
- The paper states that it does not resolve whether impairments of spontaneous behavior are driven by biological pathways and mechanisms of cognitive decline, nor whether outcomes could represent biomarkers of cognitive dysfunction.
- The paper states that other neurodegenerative models were not tested to assess disease selectivity of motif abnormalities; it also notes that the ML pipeline currently has no clinical application and conclusions apply solely to mouse models of AD.
Model-written summary. Check it against the paper before relying on it.
Integrity signals
Registry
Cites
Statistics
Claims the paper makes
54 of 60 cited sources checked · 13 supported · 0 flagged · accuracy not yet measured
- Premisecitation
In patients and mouse models of AD, fibrinogen extravasates in the brain at sites of impaired blood-brain barrier integrity and cerebrovascular dysfunction, which are early features of AD pathogenesis.
“In patients and mouse models of AD, fibrinogen extravasates in the brain at sites of impaired blood-brain barrier integrity and cerebrovascular dysfunction, which are early features of AD pathogenesis.21,39-42”
Quote in the paperSubstantial claim5 of 5 sources checked- cites [21] Merlini, 2019, Fibrinogen induces microglia-mediated spine elimination and cognitive impairment in an Alzheimer's disease model, Neuron source
- cites [39] Iturria-Medina, 2016, Early role of vascular dysregulation on late-onset Alzheimer's disease based on multifactorial data-driven analysis, Nat. Commun. source
The Discussion (page not provided) says the paper’s results suggest vascular dysregulation is an early event in late-onset AD and notes evidence of blood–brain barrier permeability breakdown in the human hippocampus. It does not report fibrinogen extravasation at affected brain sites in patients or mouse models. (LLM, conf=0.88)
Experimental — accuracy not yet measured - cites [40] Nation, 2019, Blood-brain barrier breakdown is an early biomarker of human cognitive dysfunction, Nat. Med. source
The Introduction describes neurovascular dysfunction and BBB breakdown as early features of AD. It also reports that CSF fibrinogen correlates with a marker of BBB breakdown, as shown in Extended Data Figure 2 (no page number provided). Neither passage establishes fibrinogen extravasation at impaired brain sites in AD patients or mouse models. (LLM, conf=0.87)
Experimental — accuracy not yet measured - cites [41] Iadecola, 2013, The pathobiology of vascular dementia, Neuron source
The sections on white matter injury describe fibrinogen extravasation following BBB disruption, while the section on overlap with neurodegenerative dementia describes early cerebrovascular dysfunction in AD and BBB disruption associated with ApoE4. The paper does not establish that fibrinogen extravasates specifically in patients and mouse models of AD at those sites; its fibrinogen discussion concerns vascular white matter injury. (LLM, conf=0.87)
Experimental — accuracy not yet measured - cites [42] Ryu, 2009, A leaky blood-brain barrier, fibrinogen infiltration and microglial reactivity in inflamed Alzheimer's disease brain, J. Cell Mol. Med. source
The Introduction reports fibrinogen extravasation through a weakened blood–brain barrier in AD brains and fibrin deposition with damaged vasculature in transgenic mouse models. The Discussion describes extravascular fibrinogen near abnormal vessels in AD tissue. Neither passage establishes that blood–brain barrier impairment or cerebrovascular dysfunction is an early feature of AD pathogenesis. (LLM, conf=0.86)
Experimental — accuracy not yet measured
- Premisecitation
Fggγ390-396A reduces AD-related pathogenesis, including amyloid burden, dystrophic neurites, synaptic loss, microgliosis, and neuroinflammation in 5xFAD-Fggγ390-396A mice relative to 5xFAD mice.
“Fggγ390–396A reduces AD-related pathogenesis, including amyloid burden, dystrophic neurites, synaptic loss, microgliosis, and neuroinflammation in 5xFAD-Fggγ390–396A mice relative to 5xFAD mice.21,22,34”
Quote in the paperSubstantial claim3 of 3 sources checked- cites [21] Merlini, 2019, Fibrinogen induces microglia-mediated spine elimination and cognitive impairment in an Alzheimer's disease model, Neuron source
- cites [22] Mendiola, 2023, Defining blood-induced microglia functions in neurodegeneration through multiomic profiling, Nat. Immunol. source
In Results, “Fibrin drives neurotoxic microglia programs in AD mice” (page unavailable), the paper reports reduced neurodegeneration, a reduced neurodegenerative microglial gene signature, and fewer oxidative-stress-producing microglia around plaques in 5XFAD:Fggγ390–396A mice. It does not report reductions in amyloid burden, dystrophic neurites, synaptic loss, or overall microgliosis in that comparison; the reported microglial changes support only part of the claimed neuroinflammation reduction. (LLM, conf=0.86)
Experimental — accuracy not yet measured - cites [34] Ryu, 2018, Fibrin-targeting immunotherapy protects against neuroinflammation and neurodegeneration, Nat. Immunol. source
The Results section ('Fibrin immunotherapy protects AD mice'; page unavailable) reports that the 5B8 antibody reduced microglial activation and inflammatory gene expression in 5XFAD mice. It does not report a comparison of 5XFAD-Fggγ390–396A mice with 5XFAD mice, or reductions in dystrophic neurites or synaptic loss; it says 5B8 did not significantly affect amyloid plaques. (LLM, conf=0.85)
Experimental — accuracy not yet measured
- Premisecitation
The binding of Aβ to fibrinogen inhibits fibrin degradation, resulting in chronic fibrin deposition in the AD brain.
“The binding of Aβ to fibrinogen inhibits fibrin degradation, resulting in chronic fibrin deposition in the AD brain.32”
Quote in the paperSubstantial claim1 of 1 sources checked- cites [32] Cortes-Canteli, 2010, Fibrinogen and beta-amyloid association alters thrombosis and fibrinolysis: a possible contributing factor to Alzheimer's disease, Neuron source
The Discussion ('A possible new role for Aβ in Alzheimer's disease') says that Aβ–fibrinogen association produces clots more resistant to lysis. It says this could explain fibrin(ogen) accumulation in plaques, while the 'Hypothetical model' presents fibrin persistence in the AD brain as a proposed mechanism. The claim overstates that proposed link as an established cause of chronic deposition. (LLM, conf=0.86)
Experimental — accuracy not yet measured
- Interpretationcitation
These behavioral changes may be the earliest symptoms of preclinical stages of AD.
“Thus, these behavioral changes may be the earliest symptoms of preclinical stages of AD.13,14”
Quote in the paperSubstantial claim2 of 2 sources checked- cites [13] Rosenberg, 2013, The association of neuropsychiatric symptoms in MCI with incident dementia and Alzheimer disease, Am. J. Geriatr. Psychiatr. source
- cites [14] Peters, 2013, Neuropsychiatric symptoms as risk factors for progression from CIND to dementia: the Cache County Study, Am. J. Geriatr. Psychiatr. source
The Discussion (page unavailable) reports that neuropsychiatric symptoms, even at low severity, increased the risk of progression from cognitive impairment without dementia (CIND) to dementia. The Results (page unavailable) specifically identify nighttime behaviors as a risk factor for progression to AD. The paper does not establish that behavioral changes are the earliest symptoms of preclinical AD; its participants already had cognitive impairment. (LLM, conf=0.82)
Experimental — accuracy not yet measured
- Findingown data
Kinematics analyses of egocentric coordinates of the 30 motifs revealed high specificity regardless of sex and genotype.
“Kinematics analyses of egocentric coordinates of the 30 motifs also revealed high specificity regardless of sex and genotype (Figures 2G, S1C, and S2), enabling us to precisely study mouse behavior.”
Quote in the paperSubstantial claim
- Findingown data
In the hidden platform component, AppNL-G-F mice showed deficits in distance (p = 0.025) and latency (p = 0.017).
“In the hidden platform (spatial) component, AppNL-G-F mice showed deficits in distance (p = 0.025) and latency (p = 0.017) (Figure 1A).”
Quote in the paperSubstantial claim - Findingown data
Direct comparisons of control groups WT (Fgg+/+) and Fggγ390-396A mice revealed no significant differences in usage of any motifs.
“Direct comparisons of the control groups, WT (Fgg+/+), and Fggγ390–396A mice revealed no significant differences in the usage of any of the motifs (Figure S4A).”
Quote in the paperSubstantial claim - Findingown data
5xFAD mice had prominent ML behavioral alterations in 17 of 30 motifs.
“We found that 5xFAD mice had prominent ML behavioral alterations in 17 of 30 motifs (Figure 6A).”
Quote in the paperSubstantial claim - Findingown data
No sex effects were seen in 5xFAD and 5xFAD-Fggγ390-396A mice.
“No sex effects were seen in 5xFAD and 5xFAD-Fggγ390–396A mice (Figure S4B).”
Quote in the paperNot a substantial claim - Findingown data
Motifs of the same community were similarly affected by App.
“Notably, motifs of the same community were similarly affected by App (Figure 4C, top), validating the concept that communities represent cohesive behaviors.”
Quote in the paperSubstantial claim - Findingown data
In the probe trial, AppNL-G-F mice had no preference for the target quadrant.
“In the probe trial (platform removed), AppNL-G-F mice had no preference for the target quadrant (Figure 1B), indicating memory impairment.”
Quote in the paperSubstantial claim - Findingown data
6-month-old AppNL-G-F mice did not display deficits.
“While 6-month-old AppNL-G-F mice did not display deficits (Figure S1D), we found that motif usage was markedly altered in 13-month-old AppNL-G-F mice”
Quote in the paperNot a substantial claim - Findingown data
Hierarchical behavioral dendrogram showed significant 5xFAD-dependent usage alterations in eight of nine identified behavioral communities (FDR-BH, q < 0.05).
“These categories were confirmed by the hierarchical behavioral dendrogram, which showed significant 5xFAD-dependent usage alterations in eight of nine identified behavioral communities (FDR-BH, q < 0.05) (Figure 7A, red branches).”
Quote in the paperSubstantial claim - Findingown data
In AppNL-G-F mice, particularly females, behavioral sequences were less predictable than in WT mice at all orders, indicating increased entropy or randomness in behavioral organization.
“In AppNL-G-F mice, particularly females, behavioral sequences were less predictable than in WT mice at all orders, indicating increased entropy or randomness in behavioral organization.”
Quote in the paperSubstantial claim - Findingown data
Kinematic and video analyses of motifs significantly improved in 5xFAD-Fggγ390-396A mice were assigned to generic behavioral categories: run (motif 2), walkup (motifs 6, 11, and 28), groom-up (motif 17), groom-down (motifs 10 and 12), and walk (motif 27).
“Kinematic and video analyses of the motifs that were significantly improved in 5xFAD-Fggγ390–396A mice were assigned to the following generic behavioral categories: run (motif 2), walkup (motifs 6, 11, and 28), groom-up (motif 17), groom-down (motifs 10 and 12), and walk (motif 27) (Figures 6E and S5).”
Quote in the paperNot a substantial claim - Findingown data
Hierarchical clustering of motif usage and mice revealed prominent disease-associated differences.
“Hierarchical clustering of motif usage and mice also revealed prominent disease-associated differences (Figure 7C).”
Quote in the paperNot a substantial claim - Findingown data
VAME identified consistent and robust differences in motif use between male and female controls.
“Notably, VAME identified consistent and robust differences in motif use between male and female controls (Figure 3D).”
Quote in the paperSubstantial claim - Findingown data
Motif usage was markedly altered in 13-month-old AppNL-G-F mice, with significant (q < 0.05) alterations in eight of 30 identified motifs.
“we found that motif usage was markedly altered in 13-month-old AppNL-G-F mice, with significant (q < 0.05) alterations in eight of 30 identified motifs, as shown by false discovery rate test with Benjamini-Hochberg correction for multiple comparisons (FDR-BH) (Figure 3A).”
Quote in the paperSubstantial claim - Findingown data
AppNL-G-F mice showed a well-defined gradient of behavioral impairment.
“AppNL-G-F mice also showed a well-defined gradient of behavioral impairment, indicating that spontaneous behavioral changes capture interindividual variability.”
Quote in the paperSubstantial claim - Findingown data
AppNL-G-F mice's time in the target quadrant was not significantly different from WT littermate controls (p = 0.108).
“However, their time in the target quadrant was not significantly different from WT littermate controls (p = 0.108, Figure 1B).”
Quote in the paperNot a substantial claim - Findingown data
The effects of App genotype and the direction of disease-associated changes in motif use were remarkably consistent across females and males.
“However, the effects of App genotype and the direction of disease-associated changes in motif use were remarkably consistent across females and males, indicating that VAME can resolve sex and genotype interactions.”
Quote in the paperSubstantial claim - Findingown data
There were significant differences in clustering between AppNL-G-F and WT mice (p = 0.0015, Mann-Whitney rank-sum test).
“We detected significant differences in clustering between AppNL-G-F and WT mice (p = 0.0015, Mann-Whitney rank-sum test) (Figure 4D), indicating abnormalities in the overall motif use profile.”
Quote in the paperSubstantial claim - Findingown data
22-month-old AppNL-G-F mice exhibited mild learning and memory deficits in the Morris water maze relative to their wild-type littermate controls.
“22-month-old AppNL-G-F mice exhibited mild learning and memory deficits in the Morris water maze relative to their wild-type (WT) littermate controls (Figures 1A and 1B).”
Quote in the paperSubstantial claim - Findingown data
Motifs of each mouse fully overlapped with the corresponding motifs of other mice regardless of sex and genotype.
“Notably, motifs of each mouse fully overlapped with the corresponding motifs of the other mice regardless of sex and genotype (Figure S1B), indicating that motif identification was consistent throughout the cohort.”
Quote in the paperSubstantial claim - Findingown data
The k-means algorithm had a higher overall probability of classifying genotype correctly than HMM (93.8% vs. 90.6%).
“The k-mean algorithm had a higher overall probability of classifying genotype correctly than HMM (93.8% vs. 90.6%) and was used to generate a unique sequence of motifs for each mouse (Figure 2E).”
Quote in the paperNot a substantial claim - Findingown data
Nearly all 5xFAD-dependent alterations were ameliorated in 5xFAD-Fggγ390-396A mice.
“Remarkably, nearly all of these 5xFAD-dependent alterations were ameliorated in 5xFAD-Fggγ390–396A mice, indicating global restoration of spontaneous behavior, including motif use and speeds (Figures 6A and 6B).”
Quote in the paperSubstantial claim - Findingown data
Motifs were well organized into communities.
“As expected from the transition probability matrix (Figure 4B), we found that motifs were well organized into communities (Figures 5C and 5D).”
Quote in the paperNot a substantial claim - Findingown data
5xFAD-dependent abnormalities in motif transition and motif speed delta indices were ameliorated in 5xFAD-Fggγ390-396A mice.
“We found that 5xFAD-dependent abnormalities in the motif transition and motif speed delta indices were ameliorated in 5xFAD-Fggγ390–396A mice (Figures 6C and 6D).”
Quote in the paperSubstantial claim - Findingown data
The eight motifs affected in AppNL-G-F mice belong to generic behavioral categories: walk (motif 0), fast explore (motif 5), slow explore (motif 15), rear (motifs 2, 6), run (motifs 12 and 18), and sit up (motif 19).
“Kinematic and video analyses of the eight motifs affected in AppNL-G-F mice revealed that the motifs belong to the following generic behavioral categories: walk (motif 0), fast explore (motif 5), slow explore (motif 15), rear (motifs 2, 6), run (motifs 12 and 18), and sit up (motif 19) (Figures 3B, 3C, S2, and S3A).”
Quote in the paperNot a substantial claim - Findingown data
Motifs and communities were strongly modulated by time and genotype.
“We found that motifs and communities were strongly modulated by time and genotype.”
Quote in the paperSubstantial claim - Findingown data
Ambulatory behaviors were the hub around which the rest of the behaviors were arranged.
“Interestingly, ambulatory behaviors were the hub around which the rest of the behaviors were arranged.”
Quote in the paperNot a substantial claim - Findingown data
Communities displayed distinct allocentric speeds consistent with the represented behaviors.
“Although the VAME training data included only egocentric coordinates (relative to mouse center) without allocentric speed information, communities displayed distinct allocentric speeds consistent with the represented behaviors.”
Quote in the paperSubstantial claim - Findingown data
The probability matrix of motif-motif transitions consistently showed that motifs within each community formed distinct clusters.
“The probability matrix of motif-motif transitions consistently showed that motifs within each community formed distinct clusters (Figure 4B), demonstrating highly structured behavioral dynamics.”
Quote in the paperSubstantial claim - Findingown data
Only the VAME outcomes differentiated 5xFAD-Fggγ390-396A from control mice.
“Notably, only the VAME outcomes differentiated 5xFAD-Fggγ390–396A from control mice, suggesting caution in interpreting therapeutic interventions in mouse models, as conventional approaches may not fully capture remaining disease phenotypes in rescued mice.”
Quote in the paperSubstantial claim - Findingnone given
VAME captured robust differences in the spontaneous behavior of males and females.
“While sex and genotype interactions are difficult to detect and assess with conventional behavioral tests, VAME captured robust differences in the spontaneous behavior of males and females, making it an ideal platform to assess sex-genotype interactions.”
Quote in the paperSubstantial claim - Findingown data
ML, but not conventional, outcomes distinguished rescued 5xFAD-Fggγ390-396A mice from controls.
“Remarkably, ML, but not conventional, outcomes distinguished rescued 5xFAD-Fggγ390–396A mice from controls, indicating that conventional approaches lack sufficient sensitivity and specificity to capture remaining disease phenotypes in rescued mice, which may result in an overestimation of the beneficial effects of the intervention.”
Quote in the paperSubstantial claim - Findingown data
Both AppNL-G-F and 5xFAD mice displayed overlapping abnormalities, including increased motif speed, a higher number of transitions, and reduced motif dwell time.
“Interestingly, both AppNL-G-F and 5xFAD mice displayed overlapping abnormalities, including increased motif speed, a higher number of transitions, and reduced motif dwell time.”
Quote in the paperSubstantial claim - Findingown data
Motifs in the same communities responded to AppNL-G-F and 5xFAD in similar ways in both direction and magnitude.
“Indeed, motifs in the same communities responded to AppNL-G-F and 5xFAD in similar ways in both direction and magnitude, suggesting that communities cluster disease-relevant behaviors.”
Quote in the paperSubstantial claim - Findingown data
For WT vs. AppNL-G-F, the conventional behavior outcomes model correctly classified WT = 50.0% and AppNL-G-F = 77.8% (AIC = 37.052).
“(correct classification: WT = 50.0% and AppNL–G-F = 77.8%; AIC = 37.052)”
Quote in the paperSubstantial claim - Findingown data
ML outcomes classified genotypes and treatments with greater specificity and sensitivity than conventional behavior outcomes from the same behavioral videos.
“Moreover, when we assessed the specificity and sensitivity of ML (motif usage) and conventional behavior (distance, speed, and location) outcomes from the same behavioral videos, the ML outcomes classified genotypes and treatments with greater specificity and sensitivity.”
Quote in the paperSubstantial claim - Findingown data
Motifs with ambulatory kinematics (e.g., run or fast-explore) exhibited higher allocentric speeds than stationary motifs (e.g., groom or sit).
“Although VAME segmentation does not incorporate additional variables such as mouse speed, height, or location, motifs with ambulatory kinematics (e.g., run or fast-explore) exhibited higher allocentric speeds than stationary motifs (e.g., groom or sit).”
Quote in the paperSubstantial claim - Findingown data
Genetic blockade restored the impaired transition networks in 5xFAD-Fggγ390-396A mice.
“the restoration of the impaired transition networks in 5xFAD-Fggγ390–396A mice”
Quote in the paperSubstantial claim - Findingown data
There were no significant differences in motif usage between Fggγ390-396A and WT mice.
“Because there were no significant differences in motif usage between Fggγ390–396A and WT mice (Figure S4A), these two groups of mice were combined and treated as control group.”
Quote in the paperSubstantial claim - Findingown data
The same communities were also apparent in the unsupervised layout of transition networks.
“The same communities were also apparent in the unsupervised layout of transition networks, a method that relies on the overall topological connectivity among motifs and is not biased by prior community assignments.”
Quote in the paperNot a substantial claim - Findingown data
In the 5xFAD cohort, WT and Fggγ390-396A mice did not differ in motif usage.
“Since their motif usage did not differ (Figure S4A), these two groups were combined into a single control group.”
Quote in the paperNot a substantial claim - Findingown data
Transitions were particularly sensitive to AppNL-G-F and 5xFAD.
“By creating and calculating global delta indices (transitions, usage, and speed indices), we found that transitions were particularly sensitive to AppNL-G-F and 5xFAD, suggesting that the core abnormality in AD models is the disorganization of behavioral sequences rather than the total time spent in each behavior.”
Quote in the paperSubstantial claim - Findingown data
k-means demonstrated higher probability of correct genotype classification using logistic regression compared to HMM (93.8% vs. 90.6%).
“Because k-means demonstrated higher probability of correct genotype classification using logistic regression compared to HMM (93.8% vs. 90.6%) (Figure S1A), k-means was chosen to identify motifs.”
Quote in the paperSubstantial claim - Findingown data
Genetic blockade of fibrinogen-microglia interactions in 5xFAD-Fggγ390-396A mice reduced virtually all the motif usage abnormalities found in 5xFAD mice.
“Notably, we found that genetic blockade of fibrinogen-microglia interactions in 5xFAD-Fggγ390–396A mice reduced virtually all the motif usage abnormalities found in 5xFAD mice.”
Quote in the paperSubstantial claim - Findingown data
For control, Fggγ390-396A, and 5xFAD-Fggγ390-396A, the conventional behavior outcomes model correctly classified control = 90.9%, Fggγ390-396A = 69.2%, and 5xFAD-Fggγ390-396A = 50.0% (AIC = 70.857).
“(correct classification: control = 90.9%, Fggγ390–396A = 69.2%, and 5xFAD-Fggγ390–396A = 50.0%; AIC = 70.857)”
Quote in the paperSubstantial claim - Interpretationown data
AppNL-G-F mice displayed abnormal habituation and sensitization responses to the environmental context in virtually all identified behaviors.
“Thus, AppNL-G-F mice displayed abnormal habituation and sensitization responses to the environmental context in virtually all identified behaviors.”
Quote in the paperSubstantial claim - Interpretationown data
Fggγ390-396A does not change spontaneous behavior in a non-AD background.
“Thus, Fggγ390–396A does not change spontaneous behavior in a non-AD background.”
Quote in the paperSubstantial claim - Interpretationown data
AppNL-G-F mice exhibit robust age-dependent behavioral alterations in spontaneous behavior.
“Thus, AppNL-G-F mice exhibit robust age-dependent behavioral alterations in spontaneous behavior.”
Quote in the paperSubstantial claim- Rests on an assumption: The 30 postural motifs identified by VAME capture the full repertoire of spontaneous mouse behavior relevant to AD-related behavioral alterations.
- Interpretationown data
The amelioration of 5xFAD-dependent alterations in 5xFAD-Fggγ390-396A mice indicates global restoration of spontaneous behavior, including motif use and speeds.
“Remarkably, nearly all of these 5xFAD-dependent alterations were ameliorated in 5xFAD-Fggγ390–396A mice, indicating global restoration of spontaneous behavior, including motif use and speeds (Figures 6A and 6B).”
Quote in the paperSubstantial claim - Interpretationown data
Fibrinogen is a key mediator of behavioral abnormalities in 5xFAD mice.
“Together, these results indicate that fibrinogen is a key mediator of behavioral abnormalities in 5xFAD mice.”
Quote in the paperSubstantial claim- Rests on an assumption: The rescue of spontaneous behavioral alterations in 5xFAD-Fggγ390–396A mice is due to reduced AD-related neuroinflammation rather than to other effects of the Fggγ390–396A mutation.
- Interpretationown data
VAME analysis is more sensitive and specific than conventional outcomes for AppNL-G-F and 5xFAD mice.
“Thus, VAME analysis is more sensitive and specific than conventional outcomes for AppNL-G-F and 5xFAD mice.”
Quote in the paperSubstantial claim - Interpretationreasoning
Each motif represents a brief sequence of behavior that is consistently identified across mice and over time.
“Thus, each motif represents a brief sequence of behavior that is consistently identified across mice and over time.”
Quote in the paperSubstantial claim- Rests on an assumption: The 30 postural motifs identified by VAME capture the full repertoire of spontaneous mouse behavior relevant to AD-related behavioral alterations.
- Study design and conductnone given
A limitation of the study is that specific behavioral domains (e.g., cognitive functions) and functional brain topology were not explicitly assessed.
“A limitation of our study assessing spontaneous behavior is that specific behavioral domains (e.g., cognitive functions) and functional brain topology were not explicitly assessed.”
Quote in the paperNot a substantial claim - Interpretationnone given
The ability to scale ML approaches to large cohorts of mice facilitates well-powered evaluation of treatment effectiveness across sex and dosages.
“The ability to scale ML approaches to large cohorts of mice also facilitates well-powered evaluation of treatment effectiveness across sex and dosages.”
Quote in the paperSubstantial claim - Interpretationreasoning
Fibrinogen is a key driver of the disorganization of behavioral sequences in 5xFAD mice.
“These findings, coupled with the restoration of the impaired transition networks in 5xFAD-Fggγ390–396A mice, suggest that fibrinogen is a key driver of the disorganization of behavioral sequences in 5xFAD mice.”
Quote in the paperSubstantial claim - Interpretationown data
VAME is a high-throughput approach for unbiased assessment of therapeutic interventions.
“Our findings show that VAME is a high-throughput approach for unbiased assessment of therapeutic interventions.”
Quote in the paperSubstantial claim - Interpretationreasoning
Conventional approaches may result in an overestimation of the beneficial effects of the intervention.
“which may result in an overestimation of the beneficial effects of the intervention”
Quote in the paperSubstantial claim - Interpretationreasoning
The resulting communities represent cohesive postural sequences of ethologically relevant behavior.
“The resulting communities represent cohesive postural sequences of ethologically relevant behavior.”
Quote in the paperSubstantial claim - Interpretationown data
These results suggest that these distinct mouse models share common mechanisms underlying AD-related pathogenesis.
“These results suggest that these distinct mouse models share common mechanisms underlying AD-related pathogenesis.”
Quote in the paperSubstantial claim - Interpretationnone given
ML approaches to assess spontaneous behavior provide a direct, sensitive, and unbiased measure of brain dysfunction induced by AD-related mechanisms in mouse models.
“Our findings suggest that ML approaches to assess spontaneous behavior provide a direct, sensitive, and unbiased measure of brain dysfunction induced by AD-related mechanisms in mouse models.”
Quote in the paperSubstantial claim- Rests on an assumption: The behavioral abnormalities detected by VAME in mouse models are caused by AD-related brain dysfunction rather than peripheral or motor differences.
- Interpretationown data
Kinematics reveal ethologically-relevant behaviors that are not obvious to human observers.
“which reveal ethologically-relevant behaviors not obvious to human observers”
Quote in the paperSubstantial claim - Premisenone given
Clustering motifs into communities provides a hierarchical organization of mouse behavior that incorporates the probability of motif-motif transitions, thereby grouping motifs according to the likelihood of sequential use rather than pose similarity.
“Clustering motifs into communities provides a hierarchical organization of mouse behavior that incorporates the probability of motif-motif transitions, thereby grouping motifs according to the likelihood of sequential use rather than pose similarity.”
Quote in the paperSubstantial claim - Interpretationcitation
ML approaches for behavioral segmentation have the potential to detect subtle or overlooked behavioral abnormalities linked to early brain dysfunction.
“ML approaches for behavioral segmentation have the potential to detect subtle or overlooked behavioral abnormalities linked to early brain dysfunction.”
Quote in the paperSubstantial claim - Interpretationown data
VAME is an ideal platform to assess sex-genotype interactions.
“making it an ideal platform to assess sex-genotype interactions”
Quote in the paperSubstantial claim - Interpretationown data
The core abnormality in AD models is the disorganization of behavioral sequences rather than the total time spent in each behavior.
“suggesting that the core abnormality in AD models is the disorganization of behavioral sequences rather than the total time spent in each behavior”
Quote in the paperSubstantial claim - Interpretationown data
Conventional approaches lack sufficient sensitivity and specificity to capture remaining disease phenotypes in rescued mice.
“indicating that conventional approaches lack sufficient sensitivity and specificity to capture remaining disease phenotypes in rescued mice”
Quote in the paperSubstantial claim - Interpretationown data
Communities cluster disease-relevant behaviors.
“suggesting that communities cluster disease-relevant behaviors”
Quote in the paperSubstantial claim
- May combine two claims. 48
- Classified as probably not a substantial claim. 30
May combine two claims.
InterpretationcitationApp KI mice may model preclinical stages of AD, characterized by progressive amyloidosis and gliosis without prominent cognitive deterioration, tau pathology, or neurodegeneration.
“As suggested,15,20 App KI mice may model pre-clinical stages of AD, characterized by progressive amyloidosis and gliosis without prominent cognitive deterioration, tau pathology, or neurodegeneration.”
Quote in the paperSubstantial claim1 of 2 sources checkedClassified as probably not a substantial claim. May combine two claims.
PremisecitationAD begins several decades before the clinical onset of dementia with progressive amyloid accumulation preceding significant tau pathology and neurodegeneration.
“Alzheimer’s disease (AD) begins several decades before the clinical onset of dementia with progressive amyloid accumulation preceding significant tau pathology and neurodegeneration.10,11”
Quote in the paperSubstantial claim2 of 2 sources checked- cites [10] Hanseeuw, 2019, Association of amyloid and tau with cognition in preclinical Alzheimer disease: A longitudinal study, JAMA Neurol. source
The Discussion reports that amyloid accumulation preceded changes in inferior temporal tau and cites prior evidence that amyloid positivity occurs many years before dementia onset. It does not establish a several-decade interval, and the Limitations section says the sequence of neurodegenerative changes relative to amyloid and tau remains to be determined. (LLM, conf=0.88)
Experimental — accuracy not yet measured - cites [11] Sperling, 2011, Toward defining the preclinical stages of Alzheimer's disease: Recommendations from the National Institute on Aging-Alzheimer's Association workgroups on diagnostic guidelines for Alzheimer's disease, Alzheimers Dement. source
The Introduction (page unavailable) says the AD process begins “years, if not decades” before clinical dementia, not that it consistently begins several decades earlier. Sections 2 and 4 describe amyloid accumulation as a hypothesized early event and tau pathology and neurodegeneration as downstream changes, but present that sequence as a proposed model rather than an established fact. (LLM, conf=0.88)
Experimental — accuracy not yet measured
Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.
PremisecitationThese behavioral changes are associated with increased risk of dementia and AD, suggesting that they contribute to AD pathogenesis.
“These behavioral changes are associated with increased risk of dementia and AD,13,14 suggesting that they contribute to AD pathogenesis.13,14”
Quote in the paperNot a substantial claim2 of 2 sources checked- cites [13] Rosenberg, 2013, The association of neuropsychiatric symptoms in MCI with incident dementia and Alzheimer disease, Am. J. Geriatr. Psychiatr. source
The Results and Discussion report that neuropsychiatric symptoms in people with MCI are associated with increased risk of dementia and AD. But the Discussion says the authors cannot infer, without biomarkers or autopsy data, that those symptoms reflect the AD neuropathological process. The paper therefore does not establish that the behavioral changes contribute to AD pathogenesis. (LLM, conf=0.90)
Experimental — accuracy not yet measured - cites [14] Peters, 2013, Neuropsychiatric symptoms as risk factors for progression from CIND to dementia: the Cache County Study, Am. J. Geriatr. Psychiatr. source
The Results (page unavailable) associate neuropsychiatric symptoms with progression from CIND to all-cause dementia and identify nighttime behaviors as a risk factor for AD. The Discussion (page unavailable) describes these symptoms as predictors, but does not establish that they contribute to AD pathogenesis; the aMCI subgroup also showed no significant neuropsychiatric-symptom risk factors for dementia or AD. (LLM, conf=0.88)
Experimental — accuracy not yet measured
May combine two claims.
FindingcitationSpontaneous behavioral abnormalities in 5xFAD mice were prevented by an AD-related therapeutic intervention, Fggγ390-396A, that prevents CD11b receptor-mediated interactions with microglia and reduces AD-related pathogenesis, including amyloid burden, dystrophic neurites, synaptic loss, microgliosis, and neuroinflammation.
“Third, and more importantly, spontaneous behavioral abnormalities in 5xFAD mice were prevented by an AD-related therapeutic intervention, Fggγ390–396A, that prevents CD11b receptor-mediated interactions with microglia22,33 and reduces AD-related pathogenesis, including amyloid burden, dystrophic neurites, synaptic loss, microgliosis, and neuroinflammation.21,22,34”
Quote in the paperSubstantial claim3 of 3 sources checked- cites [22] Mendiola, 2023, Defining blood-induced microglia functions in neurodegeneration through multiomic profiling, Nat. Immunol. source
Split into 2 atoms — 1 confirmed, 1 partially supported, 0 disputed, 0 misattributed, 0 unverifiable. Expand the claim to see per-atom verdicts.
Experimental — accuracy not yet measured - cites [33] Petersen, 2018, Fibrinogen in neurological diseases: mechanisms, imaging and therapeutics, Nat. Rev. Neurosci. source
Split into 2 atoms — 0 confirmed, 2 partially supported, 0 disputed, 0 misattributed, 0 unverifiable. Expand the claim to see per-atom verdicts.
Experimental — accuracy not yet measured - cites [21] Merlini, 2019, Fibrinogen induces microglia-mediated spine elimination and cognitive impairment in an Alzheimer's disease model, Neuron source
Split into 2 atoms — 1 confirmed, 1 partially supported, 0 disputed, 0 misattributed, 0 unverifiable. Expand the claim to see per-atom verdicts.
Experimental — accuracy not yet measured
May combine two claims.
PremisecitationA fibrin-targeting antibody that selectively inhibits microglia activation without adverse anticoagulant effects protects from neurodegeneration and suppresses AD-related gene networks in 5xFAD mice.
“Accordingly, a fibrin-targeting antibody that selectively inhibits microglia activation without adverse anticoagulant effects protects from neurodegeneration and suppresses AD-related gene networks in 5xFAD mice.28,34”
Quote in the paperSubstantial claim2 of 2 sources checked- cites [28] Ryu, 2024, Fibrin drives thromboinflammation and neuropathology in COVID-19, Nature source
The ‘Fibrin-targeting antibody in SARS-CoV-2’ section says 5B8 protects against amyloid-driven neurodegeneration without adverse effects on haemostasis; the Discussion also says it does not affect normal clotting time. This paper reports reduced microglial reactivity and altered gene expression in SARS-CoV-2-infected mice, but does not establish that the antibody suppresses AD-related gene networks in 5xFAD mice. (LLM, conf=0.86)
Experimental — accuracy not yet measured - cites [34] Ryu, 2018, Fibrin-targeting immunotherapy protects against neuroinflammation and neurodegeneration, Nat. Immunol. source
Probably not a substantial claim: may not pass S3 (load-bearing).
PremisecitationML approaches have been developed to predict these behavioral abnormalities in AD patients by using data on sleep patterns and physical activity collected by actigraphy from wearable devices.
“ML approaches have been developed to predict these behavioral abnormalities in AD patients by using data on sleep patterns and physical activity collected by actigraphy from wearable devices.38”
Quote in the paperNot a substantial claim1 of 1 sources checked- cites [38] Cho, 2023, Machine learning-based predictive models for the occurrence of behavioral and psychological symptoms of dementia: model development and validation, Sci. Rep. source
The Discussion says the models predicted behavioral and psychological symptoms of dementia using actigraphy measures of sleep and activity, among other inputs. But the Results (Participant characteristics) show that the study included people with several types of dementia, not only Alzheimer’s disease; the claim narrows the population to AD patients. (LLM, conf=0.88)
Experimental — accuracy not yet measured
May combine two claims.
InterpretationcitationSpontaneous behavioral abnormalities were age dependent and therefore tracked the progression of AD-related pathological and cognitive alterations.
“Second, spontaneous behavioral abnormalities were age dependent and therefore tracked the progression of AD-related pathological and cognitive alterations.16,19,23”
Quote in the paperSubstantial claim3 of 3 sources checked- cites [16] Xia, 2022, Novel App knock-in mouse model shows key features of amyloid pathology and reveals profound metabolic dysregulation of microglia, Mol. Neurodegener. source
The Results section on behavioral phenotypes reports an age-dependent increase in spontaneous activity in AppSAA knock-in mice. The Results section on amyloid pathology reports progressive plaque deposition, but the paper does not establish that the behavioral abnormalities tracked the progression of pathology or cognitive alterations; it reports habituation deficits without demonstrating that relationship. Page numbers are unavailable in our copy. (LLM, conf=0.84)
Experimental — accuracy not yet measured - cites [19] Johnson, 2020, Behavioral and neural network abnormalities in human APP transgenic mice resemble those of App knock-in mice and are modulated by familial Alzheimer's disease mutations but not by inhibition of BACE1, Mol. Neurodegener. source
The Results section, “Behavioral abnormalities observed in all three lines” (page not provided), reports behavioral and cognitive deficits in the mouse models, and the Discussion notes differences in when functional deficits began across lines. The paper does not show that spontaneous behavioral abnormalities changed with age in step with AD-related pathology and cognitive alterations; that tracking claim goes beyond its findings. (LLM, conf=0.82)
Experimental — accuracy not yet measured - cites [23] Wang, 2022, Age-Dependent Behavioral and Metabolic Assessment of App NL-G-F/NL-G-F Knock-in (KI) Mice, Front. Mol. Neurosci. source
Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.
PremisecitationMoSeq and keypoint-MoSeq use AR-HMM, whereas VAME uses deep recurrent neural networks and variational autoencoders, providing two well-established complementary approaches.
“MoSeq6 and keypoint-MoSeq35 use a statistical framework known as autoregressive hidden Markov model (AR-HMM),6,50,51 whereas VAME1,7 uses deep recurrent neural networks and variational autoencoders, providing two well-established complementary approaches.7,52,53”
Quote in the paperNot a substantial claim4 of 4 sources checked- cites [6] Wiltschko, 2015, Mapping sub-second structure in mouse behavior, Neuron source
- cites [35] Weinreb, 2024, Keypoint-MoSeq: parsing behavior by linking point tracking to pose dynamics, Nat. Methods source
- cites [1] Luxem, 2023, Open-source tools for behavioral video analysis: Setup, methods, and best practices, Elife source
In the Behavior quantification section (Table 3; page unavailable), the paper says VAME uses a variational autoencoder whose encoder and decoder are recurrent neural networks. The passages found for “MoSeq” and “autoregressive hidden Markov” do not establish that MoSeq or keypoint-MoSeq use AR-HMMs, or characterize the two as complementary approaches. (LLM, conf=0.85)
Experimental — accuracy not yet measured - cites [7] Luxem, 2022, Identifying behavioral structure from deep variational embeddings of animal motion, Commun. Biol. source
May combine two claims.
PremisecitationFibrin is present at increased levels in APOE4 AD patients, correlates with the severity of cerebral amyloid angiopathy, and is detected in amyloid-related imaging abnormalities.
“Fibrin is present at increased levels in APOE4 AD patients, correlates with the severity of cerebral amyloid angiopathy, and is detected in amyloid-related imaging abnormalities.32,43,44”
Quote in the paperNot a substantial claim2 of 3 sources checked- cites [32] Cortes-Canteli, 2010, Fibrinogen and beta-amyloid association alters thrombosis and fibrinolysis: a possible contributing factor to Alzheimer's disease, Neuron source
The Results section, “Fibrin(ogen) is deposited in CAA-positive vessels” (page not supplied), reports more vessels with fibrin(ogen) deposits in AD patients than in controls and shows deposits in CAA-positive vessels. It does not identify an APOE4 subgroup, establish a correlation with CAA severity, or report detection in amyloid-related imaging abnormalities. (LLM, conf=0.88)
Experimental — accuracy not yet measured - cites [43] Kozberg, 2022, Blood-brain barrier leakage and perivascular inflammation in cerebral amyloid angiopathy, Brain Commun. source
The Results sections ‘BBB leakage observed at all vessel grades’ and ‘BBB leakage and perivascular inflammation observed in consecutive CAA cases’ report that fibrin-positive vessels become more common with higher CAA vessel grade. ‘Grade 4 vessel identified in confirmed ARIA-E case’ reports fibrin in a vessel from one patient who developed ARIA-E. The copy does not establish increased fibrin levels specifically in APOE4 Alzheimer’s patients, and its ARIA finding is from tissue staining, not fibrin detected in imaging abnormalities. Page numbers were not provided. (LLM, conf=0.86)
Experimental — accuracy not yet measured - cites [44] Hultman, 2013, The APOE ϵ4/ϵ4 genotype potentiates vascular fibrin(ogen) deposition in amyloid-laden vessels in the brains of Alzheimer's disease patients, J. Cerebr. Blood Flow Metabol. source
Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.
PremisecitationPreviously, we described the VAME ML architecture and identified motif usage alterations in APP/PS1 transgenic mice.
“Previously, we described the VAME ML architecture and identified motif usage alterations in APP/PS1 transgenic mice.7”
Quote in the paperNot a substantial claim1 of 1 sources checked- cites [7] Luxem, 2022, Identifying behavioral structure from deep variational embeddings of animal motion, Commun. Biol. source
May combine two claims.
PremisecitationFibrinogen elevation in plasma and cerebrospinal fluid is an early biomarker of dementia risk and is associated with progression of cognitive decline in AD.
“Biomarker studies further support the notion that fibrinogen elevation in plasma and cerebrospinal fluid is an early biomarker of dementia risk and is associated with progression of cognitive decline in AD.39,45-48”
Quote in the paperNot a substantial claim4 of 5 sources checked- cites [39] Iturria-Medina, 2016, Early role of vascular dysregulation on late-onset Alzheimer's disease based on multifactorial data-driven analysis, Nat. Commun. source
The Results section ('Peripheral vascular and inflammatory alterations'; page unavailable) discusses plasma and CSF proteins as potential early biomarkers, but identifies other analytes rather than fibrinogen. Searches for fibrinogen and for elevated plasma or CSF fibrinogen in relation to dementia risk or cognitive decline found no passage addressing the specific claim. (LLM, conf=0.75)
Experimental — accuracy not yet measured - cites [45] Xu, 2008, Plasma fibrinogen is associated with cognitive decline and risk for dementia in patients with mild cognitive impairment, Int. J. Clin. Pract. source
The Results and Conclusion (page numbers unavailable) associate baseline plasma fibrinogen with subsequent cognitive decline and dementia risk in patients with mild cognitive impairment. The paper does not assess cerebrospinal fluid fibrinogen or progression of cognitive decline in patients with established Alzheimer’s disease. Its observed increase in risk for dementia of Alzheimer’s type did not reach statistical significance. (LLM, conf=0.90)
Experimental — accuracy not yet measured - cites [46] van Oijen, 2005, Fibrinogen is associated with an increased risk of Alzheimer disease and vascular dementia, Stroke source
- cites [47] Fan, 2020, The correlations between plasma fibrinogen with amyloid-beta and tau levels in patients with Alzheimer's disease, Front. Neurosci. source
The Discussion (page unavailable) reports that cognitive function decreases as plasma fibrinogen rises in AD patients, citing earlier studies. But this study measured plasma fibrinogen—not CSF fibrinogen—and does not establish it as an early biomarker of dementia risk. The Discussion explicitly says its cross-sectional design cannot determine fibrinogen’s effect on AD progression. (LLM, conf=0.88)
Experimental — accuracy not yet measured - cites [48] Yang, 2014, Prognostic polypeptide blood plasma biomarkers of Alzheimer's disease progression, J. Alzheimers Dis. source
The “Top proteome” blood analysis section reports elevated plasma fibrinogen-chain proteins in patients with progressive, rather than stable, mild cognitive impairment, supporting a link to AD progression. The Participants and prognostic-model sections describe prediction of AD onset after follow-up. The paper does not report elevated fibrinogen in cerebrospinal fluid or establish fibrinogen as a biomarker of dementia risk or measured cognitive decline; its introduction identifies different CSF biomarkers. (LLM, conf=0.88)
Experimental — accuracy not yet measured
May combine two claims.
PremisecitationFibrinogen is deposited in the brain of patients with AD and is a key mediator of neurotoxic microglia polarization and cognitive impairment in mouse models of AD.
“The blood coagulation factor fibrinogen is deposited in the brain of patients with AD and is a key mediator of neurotoxic microglia polarization and cognitive impairment in mouse models of AD.21,22,32,33”
Quote in the paperSubstantial claim3 of 3 sources checked- cites [21] Merlini, 2019, Fibrinogen induces microglia-mediated spine elimination and cognitive impairment in an Alzheimer's disease model, Neuron source
- cites [22] Mendiola, 2023, Defining blood-induced microglia functions in neurodegeneration through multiomic profiling, Nat. Immunol. source
The Results section on AD mice reports that disrupting fibrin–CD11b signaling reduces neurodegenerative microglial signatures; it also states that the mice are protected from cognitive impairment. The Discussion likewise reports protection from cognitive impairment in AD models. The passages from our copy do not establish the separate claim that fibrinogen is deposited in the brains of patients with AD. (LLM, conf=0.86)
Experimental — accuracy not yet measured - cites [32] Cortes-Canteli, 2010, Fibrinogen and beta-amyloid association alters thrombosis and fibrinolysis: a possible contributing factor to Alzheimer's disease, Neuron source
The Results section on CAA-positive vessels reports fibrin(ogen) deposits in brain tissue from patients with AD. The Results section on memory and cognitive performance reports improved spatial memory in AD mice after fibrinogen reduction. The Discussion mentions reduced microgliosis when fibrinogen is depleted, but does not establish that fibrinogen mediates neurotoxic microglial polarization. (LLM, conf=0.88)
Experimental — accuracy not yet measured
Probably not a substantial claim: may not pass S3 (load-bearing).
PremisecitationFibrin-targeting immunotherapy is currently in phase 1 clinical trials.
“Fibrin-targeting immunotherapy is currently in phase 1 clinical trials.49”
Quote in the paperNot a substantial claim1 of 1 sources checked- cites [49] Kantor, 2023, Fibrin-targeting immunotherapy for dementia, J. Prev. Alzheimers Dis. source
May combine two claims.
InterpretationcitationAD-related spontaneous behavioral alterations do not require APP overexpression and are likely linked to the pathologically elevated levels of amyloid in AppNL-G-F mice.
“These results also indicate that AD-related spontaneous behavioral alterations do not require APP overexpression and are likely linked to the pathologically elevated levels of amyloid in AppNL-G-F mice.16,19”
Quote in the paperSubstantial claim2 of 2 sources checked- cites [16] Xia, 2022, Novel App knock-in mouse model shows key features of amyloid pathology and reveals profound metabolic dysregulation of microglia, Mol. Neurodegener. source
The Results section on behavioral phenotypes says that Aβ accumulation without APP overexpression can trigger brain dysfunction in App knock-in models. The Discussion reports spontaneous-activity changes without APP overexpression in this paper’s AppSAA mice and describes amyloid pathology as a likely driver in knock-in models. It does not establish that the behavioral changes in AppNL-G-F mice are linked to their elevated amyloid levels. (LLM, conf=0.85)
Experimental — accuracy not yet measured - cites [19] Johnson, 2020, Behavioral and neural network abnormalities in human APP transgenic mice resemble those of App knock-in mice and are modulated by familial Alzheimer's disease mutations but not by inhibition of BACE1, Mol. Neurodegener. source
The Results section (‘Behavioral abnormalities observed in all three lines’) reports altered elevated-plus-maze behavior in knock-in mice, supporting the point that APP overexpression is not required. It raises Aβ or C99 as possible explanations, while the Discussion says their contributions cannot be distinguished; it does not specifically link the behavior to elevated amyloid. (LLM, conf=0.88)
Experimental — accuracy not yet measured
May combine two claims.
PremisecitationDuring the extended preclinical period prior to dementia diagnosis, many patients develop behavioral or neuropsychiatric changes, including agitation, irritability, decreased motivation, loss of empathy, and depression, which increase the risk of transition to mild cognitive impairment and AD.
“During this extended preclinical period prior to the dementia diagnosis, many patients develop behavioral or neuropsychiatric changes, including agitation, irritability, decreased motivation, loss of empathy, and depression, which increase the risk of transition to mild cognitive impairment and AD.12-14”
Quote in the paperSubstantial claim3 of 3 sources checked- cites [12] McKhann, 2011, The diagnosis of dementia due to Alzheimer's disease: Recommendations from the National Institute on Aging-Alzheimer's Association workgroups on diagnostic guidelines for Alzheimer's disease, Alzheimers Dement. source
Split into 2 atoms — 0 confirmed, 2 partially supported, 0 disputed, 0 misattributed, 0 unverifiable. Expand the claim to see per-atom verdicts.
Experimental — accuracy not yet measured - cites [13] Rosenberg, 2013, The association of neuropsychiatric symptoms in MCI with incident dementia and Alzheimer disease, Am. J. Geriatr. Psychiatr. source
Split into 2 atoms — 0 confirmed, 2 partially supported, 0 disputed, 0 misattributed, 0 unverifiable. Expand the claim to see per-atom verdicts.
Experimental — accuracy not yet measured - cites [14] Peters, 2013, Neuropsychiatric symptoms as risk factors for progression from CIND to dementia: the Cache County Study, Am. J. Geriatr. Psychiatr. source
Split into 2 atoms — 0 confirmed, 2 partially supported, 0 disputed, 0 misattributed, 0 unverifiable. Expand the claim to see per-atom verdicts.
Experimental — accuracy not yet measured
May combine two claims.
PremisecitationHumanized App knockin (KI) and APP transgenic mouse models mimic certain aspects of AD pathogenesis, including progressive amyloidosis, gliosis, and neuroinflammation, as well as network hyperexcitability and cognitive impairment, albeit to different degrees of severity across models and paradigms.
“Humanized App knockin (KI) and APP transgenic mouse models mimic certain aspects of AD pathogenesis, including progressive amyloidosis, gliosis, and neuroinflammation, as well as network hyperexcitability and cognitive impairment, albeit to different degrees of severity across models and paradigms.15-27”
Quote in the paperSubstantial claim8 of 10 sources checked- cites [15] Saito, 2014, Single App knock-in mouse models of Alzheimer’s disease, Nat. Neurosci. source
- cites [16] Xia, 2022, Novel App knock-in mouse model shows key features of amyloid pathology and reveals profound metabolic dysregulation of microglia, Mol. Neurodegener. source
The Background and Discussion (page numbers unavailable) support progressive amyloid deposition and glial responses in App knock-in models, while the Results discuss learning and memory deficits reported in APP transgenic models. The paper does not establish network hyperexcitability in its AppSAA model; altered glutamate transport and locomotor hyperactivity are not the same finding. (LLM, conf=0.86)
Experimental — accuracy not yet measured - cites [17] Latif-Hernandez, 2019, Subtle behavioral changes and increased prefrontal-hippocampal network synchronicity in APP(NL-G-F) mice before prominent plaque deposition, Behav. Brain Res. source
The Discussion reports amyloid plaque deposition and neuroinflammation in APP NL-G-F mice, and contrasts their mild behavioral phenotype with more severe impairments reported in some APP-overexpressing models. Section 3.6 reports increased prefrontal-network synchrony at 3 months, but the paper does not establish that this is network hyperexcitability; the passages also do not address gliosis. The broad list of features is therefore only partly supported by this source. (LLM, conf=0.86)
Experimental — accuracy not yet measured - cites [18] Whyte, 2018, Reduction in open field activity in the absence of memory deficits in the App(NL-G-F) knock-in mouse model of Alzheimer's disease, Behav. Brain Res. source
- cites [19] Johnson, 2020, Behavioral and neural network abnormalities in human APP transgenic mice resemble those of App knock-in mice and are modulated by familial Alzheimer's disease mutations but not by inhibition of BACE1, Mol. Neurodegener. source
The Discussion reports abnormal Aβ accumulation, epileptiform activity, and learning and memory deficits in KI and J20 mice; the Results describe different degrees of Aβ deposition across models. The Results also report plaque-associated microgliosis in J20 mice, but do not establish gliosis or neuroinflammation in both model types, or progressive amyloidosis across them; notably, I5 mice had no detectable Aβ deposition at 9–10 months. (LLM, conf=0.86)
Experimental — accuracy not yet measured - cites [20] Sasaguri, 2017, APP mouse models for Alzheimer’s disease preclinical studies, EMBO J. source
The Second-generation mouse models section (page unavailable) reports Aβ deposition with astrocyte and microglial infiltration in App knock-in mice, along with age- and model-dependent cognitive deficits. The First-generation mouse models section describes Aβ-associated neuroinflammation and behavioral phenotypes in APP transgenic mice. The review discusses epilepsy-related findings in a transgenic line but does not establish network hyperexcitability as a feature of both model types; it notes that a related channel change was not observed in the knock-in mice studied. (LLM, conf=0.82)
Experimental — accuracy not yet measured - cites [21] Merlini, 2019, Fibrinogen induces microglia-mediated spine elimination and cognitive impairment in an Alzheimer's disease model, Neuron source
The Results (page not available in our copy) identify the APP transgenic 5XFAD and hAPP-J20 models and report Aβ deposition, microglial activation, and learning and memory impairment in 5XFAD mice. The paper does not establish the claim for humanized App knock-in models or address network hyperexcitability; its reported age-related increase concerns fibrinogen deposits, not progressive amyloidosis. (LLM, conf=0.86)
Experimental — accuracy not yet measured - cites [22] Mendiola, 2023, Defining blood-induced microglia functions in neurodegeneration through multiomic profiling, Nat. Immunol. source
Split into 2 atoms — 0 confirmed, 2 partially supported, 0 disputed, 0 misattributed, 0 unverifiable. Expand the claim to see per-atom verdicts.
Experimental — accuracy not yet measured - cites [23] Wang, 2022, Age-Dependent Behavioral and Metabolic Assessment of App NL-G-F/NL-G-F Knock-in (KI) Mice, Front. Mol. Neurosci. source
Split into 2 atoms — 0 confirmed, 2 partially supported, 0 disputed, 0 misattributed, 0 unverifiable. Expand the claim to see per-atom verdicts.
Experimental — accuracy not yet measured - cites [24] Verret, 2012, Inhibitory interneuron deficit links altered network activity and cognitive dysfunction in Alzheimer model, Cell source
Split into 2 atoms — 0 confirmed, 2 partially supported, 0 disputed, 0 misattributed, 0 unverifiable. Expand the claim to see per-atom verdicts.
Experimental — accuracy not yet measured
Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.
PremisecitationMoSeq has uncovered previously undetected patterns of spontaneous behavior associated with epileptic activity, providing a behavioral biomarker of brain dysfunction.
“For example, MoSeq has uncovered previously undetected patterns of spontaneous behavior associated with epileptic activity, providing a behavioral biomarker of brain dysfunction.9”
Quote in the paperNot a substantial claim1 of 1 sources checked- cites [9] Gschwind, 2023, Hidden behavioral fingerprints in epilepsy, Neuron source
May combine two claims.
Findingown dataFemale, but not male, AppNL-G-F mice moved faster than sex-matched WT controls, particularly in the communities walk, run, and fast and slow-explore.
“Female, but not male, AppNL-G-F mice moved faster than sex-matched WT controls, particularly in the communities walk, run, and fast and slow-explore (Figure 4C, bottom).”
Quote in the paperNot a substantial claimMay combine two claims.
Findingown dataFor all genotypes, VAME outcomes for the 5xFAD cohort also had better sensitivity and specificity than conventional outcomes, categorizing 100% of mice correctly.
“Similarly, for all genotypes, VAME outcomes for the 5xFAD cohort also had better sensitivity and specificity than conventional outcomes, categorizing 100% of mice correctly (Figure 6F).”
Quote in the paperSubstantial claimMay combine two claims.
Findingown data5xFAD mice had marked abnormalities in the probability of most transitions, and the transition network was characterized by increased transitions between motifs and reduced probability of dwelling in the same motif (self-transitions).
“5xFAD mice had marked abnormalities in the probability of most transitions, and the transition network was characterized by increased transitions between motifs and reduced probability of dwelling in the same motif (self-transitions).”
Quote in the paperSubstantial claimMay combine two claims.
Findingown dataGenotype classifier using VAME outcomes had a specificity (WT classified as WT) of 92.9% and a sensitivity (AppNL-G-F classified as AppNL-G-F) of 83.3%, whereas the genotype classifier using conventional OF outcomes had much lower specificity (50.0%) and sensitivity (77.8%).
“The genotype classifier using VAME outcomes had a specificity (WT classified as WT) of 92.9% and a sensitivity (AppNL-G-F classified as AppNL-G-F) of 83.3%, whereas the genotype classifier using conventional OF outcomes had much lower specificity (50.0%) and sensitivity (77.8%) (Figure 5G).”
Quote in the paperSubstantial claimMay combine two claims.
Findingown dataAppNL-G-F mice had significant abnormalities in delta indices of motif transitions and motif use, and there was a trend toward impaired motif speed delta index.
“We found that AppNL-G-F mice had significant abnormalities in delta indices of motif transitions and motif use, and there was a trend toward impaired motif speed delta index (Figures 5E and 5F).”
Quote in the paperSubstantial claimMay combine two claims.
Findingown dataHierarchical clustering of motif-motif transitions revealed a distinct hierarchical organization among the 30 motifs, which was used to cluster the motifs into 11 behavioral communities.
“These analyses revealed a distinct hierarchical organization among the 30 motifs, which was used to cluster the motifs into 11 behavioral communities (Figures 4A and S2) according to total usage and the likelihood of sequential use (as opposed to pose similarity).”
Quote in the paperSubstantial claimMay combine two claims.
Findingown dataAppNL-G-F mice had substantial increases in the probability of transitions, particularly between motifs of different communities, resulting in decreased dwell time in ambulatory hub motifs.
“AppNL-G-F mice had substantial increases in the probability of transitions, particularly between motifs of different communities, resulting in decreased dwell time in ambulatory hub motifs (Figure 5C).”
Quote in the paperSubstantial claimMay combine two claims.
Findingown dataMost of the 5xFAD-dependent alterations in transitions were rescued or greatly mitigated in 5xFAD-Fggγ390-396A mice, indicating restoration of behavioral sequences.
“We found that most of the 5xFAD-dependent alterations in transitions were rescued or greatly mitigated in 5xFAD-Fggγ390–396A mice (Figure 7D), indicating restoration of behavioral sequences.”
Quote in the paperSubstantial claimMay combine two claims.
Findingown dataCommunities that show habituation in WT mice (B-walk, C-run, and E-fast-explore) were underrepresented in AppNL-G-F mice, whereas communities that showed sensitization in WT mice (AIV-rear, F-slow-explore, and H-sit-up) were overrepresented in AppNL-G-F mice.
“Interestingly, communities that show habituation (decreased use over time) in WT mice (B-walk, C-run, and E-fast-explore) were underrepresented in AppNL-G-F mice, whereas communities that showed sensitization (increased use over time) in WT mice (AIV-rear, F-slow-explore, and H-sit-up) were overrepresented in AppNL-G-F mice (Figure 4E).”
Quote in the paperSubstantial claimMay combine two claims.
Findingown dataAppNL-G-F mice prematurely transitioned from ambulatory communities (fast exploration or running) to slower exploratory behaviors (walk-up, sit-up, or slow exploration) and were more likely to engage in this behavioral sequence than WT controls.
“Notably, AppNL-G-F mice prematurely transitioned from ambulatory communities (fast exploration or running) to slower exploratory behaviors (walk-up, sit-up, or slow exploration) and were more likely to engage in this behavioral sequence than WT controls (Figure 5D).”
Quote in the paperSubstantial claimMay combine two claims.
Findingown dataIn a representative selection of 5xFAD-affected motifs, motif use over time was strongly modulated by genotype, and 5xFAD-Fggγ390-396A mice largely showed full or partial recovery to performances seen in control mice (p < 0.05, two-way ANOVA).
“Furthermore, in a representative selection of 5xFAD-affected motifs (Figure 7B), motif use over time was strongly modulated by genotype, and 5xFAD-Fggγ390–396A mice largely showed full or partial recovery to performances seen in control mice (p < 0.05, two-way ANOVA).”
Quote in the paperSubstantial claimMay combine two claims.
Findingown dataEach motif had a well-delimited UMAP location with smooth transitions between motifs.
“We found that each motif had a well-delimited UMAP location with smooth transitions between motifs (Figure 2F), indicating specificity of motif features and continuity of behavioral postural motion across motifs.”
Quote in the paperSubstantial claimMay combine two claims.
Findingcitation22-month-old AppNL-G-F mice had robust AD-related pathology, including amyloid deposition and neuroinflammation.
“As expected,15 22-month-old AppNL-G-F mice had robust AD-related pathology, including amyloid deposition and neuroinflammation (Figures 1C and 1D).”
Quote in the paperSubstantial claim0 of 1 sources checked- cites [15] Saito, 2014, Single App knock-in mouse models of Alzheimer’s disease, Nat. Neurosci. source
May combine two claims.
Findingown dataThe union of the set of transitions was significantly affected by 5xFAD-affected motifs and significantly restored in 5xFAD-Fggγ390-396A mice.
“We found the union of the set of transitions was significantly affected by 5xFAD-affected motifs and significantly restored in 5xFAD-Fggγ390–396A mice.”
Quote in the paperSubstantial claimMay combine two claims.
Findingown dataFor control, Fggγ390-396A, and 5xFAD-Fggγ390-396A, the ML outcomes model correctly classified control = 100%, Fggγ390-396A = 100%, and 5xFAD-Fggγ390-396A = 100% (AIC = 24.002).
“(correct classification: control = 100%, Fggγ390–396A = 100%, and 5xFAD-Fggγ390–396A = 100%; AIC = 24.002)”
Quote in the paperSubstantial claimProbably not a substantial claim: may not pass S3 (load-bearing).
Findingown dataFor control, Fggγ390-396A, and 5xFAD-Fggγ390-396A, the conventional behavior outcomes model included intercept and middle distance, wall speed, and center inactive (entry order).
“conventional behavior outcomes model included intercept and middle distance, wall speed, and center inactive (entry order)”
Quote in the paperNot a substantial claimProbably not a substantial claim: may not pass S3 (load-bearing).
Findingown dataFor WT vs. AppNL-G-F, the ML outcomes logistic regression model included constant and motifs 6 and 28 (entry order).
“For WT vs. AppNL–G-F: ML outcomes the model included constant and motifs 6 and 28 (entry order)”
Quote in the paperNot a substantial claimMay combine two claims.
Findingown dataFor WT vs. AppNL-G-F, the ML outcomes model correctly classified WT = 92.9% and AppNL-G-F = 83.3% (AIC = 24.652).
“(correct classification: WT = 92.9% and AppNL–G-F = 83.3%; AIC = 24.652)”
Quote in the paperSubstantial claimProbably not a substantial claim: may not pass S3 (load-bearing).
Findingown dataFor AppNL-G-F mice, 11 communities were identified.
“For AppNL–G-F mice, we defined the 11 identified communities as described below:”
Quote in the paperNot a substantial claimMay combine two claims.
Findingnone givenIn a side-by-side comparison of VAME and keypoint-MoSeq, these two approaches differed in motif/syllable distributions but both identified robust behavioral alterations in AppNL-G-F mice.
“In a side-by-side comparison of VAME and keypoint-MoSeq, these two approaches for behavioral segmentation differed in motif/syllable distributions but both identified robust behavioral alterations in AppNL-G-F mice.”
Quote in the paperNot a substantial claimProbably not a substantial claim: may not pass S3 (load-bearing).
FindingreasoningThe motifs in Figure S7 are different from those in the rest of the manuscript.
“Thus, the motifs in Figure S7 are different from those in the rest of the manuscript.”
Quote in the paperNot a substantial claimProbably not a substantial claim: may not pass S3 (load-bearing).
FindingreasoningMotifs and communities expressed by the respective control/WT animals differ across the two VAME models.
“Therefore, motifs and communities expressed by the respective control/WT animals differ across the two VAME models.”
Quote in the paperNot a substantial claimProbably not a substantial claim: may not pass S3 (load-bearing).
Findingown dataFor control, Fggγ390-396A, and 5xFAD-Fggγ390-396A, the ML outcomes model included intercept and motifs 2, 28, 22, 4, and 11 (entry order).
“For control, Fggγ390–396A, and 5xFAD-Fggγ390–396A: ML outcomes the model included intercept and motifs 2, 28, 22, 4, and 11 (entry order)”
Quote in the paperNot a substantial claimProbably not a substantial claim: may not pass S3 (load-bearing).
Findingnone givenVAME community videos contain some flexibility in behavioral expression.
“We therefore note that VAME community videos contain some flexibility in behavioral expression.”
Quote in the paperNot a substantial claimProbably not a substantial claim: may not pass S3 (load-bearing).
Findingown dataFor WT vs. AppNL-G-F, the conventional behavior outcomes model included middle inactive.
“conventional behavior outcomes model included middle inactive”
Quote in the paperNot a substantial claimProbably not a substantial claim: may not pass S3 (load-bearing).
Findingown dataFor k-means, the model included constant and motifs 17 and 7 (entry order).
“k-means model included constant and motifs 17 and 7 (entry order)”
Quote in the paperNot a substantial claimProbably not a substantial claim: may not pass S3 (load-bearing).
Findingown dataFor HMM, the model included constant and motifs 9, 25, and 10 (entry order).
“HMM model included constant and motifs 9, 25, and 10 (entry order)”
Quote in the paperNot a substantial claimMay combine two claims.
Findingown dataFor HMM, correct classification: WT = 92.9% and AppNL-G-F = 88.9%; overall, 90.6%; AIC = 20.029.
“(correct classification: WT = 92.9% and AppNL–G-F = 88.9%; overall, 90.6%; AIC = 20.029)”
Quote in the paperSubstantial claimProbably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.
Findingown dataFor k-means, correct classification: WT = 94.4% and AppNL-G-F = 92.8%; overall, 93.8%; AIC = 34.749.
“(correct classification: WT = 94.4% and AppNL–G-F = 92.8%; overall, 93.8%; AIC = 34.749)”
Quote in the paperNot a substantial claimMay combine two claims.
Interpretationown dataGlobal restoration of spontaneous behavior shows that fibrinogen contributes to AD-related behavioral alterations and that ML alterations are susceptible to therapeutic intervention.
“Overall, this global restoration of spontaneous behavior shows that fibrinogen contributes to AD-related behavioral alterations and that ML alterations are susceptible to therapeutic intervention.”
Quote in the paperSubstantial claim- Rests on an assumption: The rescue of spontaneous behavioral alterations in 5xFAD-Fggγ390–396A mice is due to reduced AD-related neuroinflammation rather than to other effects of the Fggγ390–396A mutation.
May combine two claims.
Interpretationown dataVAME can identify postural patterns in motion time series without supervised annotation, permitting reliable and unbiased behavioral segmentation.
“Thus, VAME can identify postural patterns in motion time series without supervised annotation, permitting reliable and unbiased behavioral segmentation.”
Quote in the paperSubstantial claim- Rests on an assumption: The 30 postural motifs identified by VAME capture the full repertoire of spontaneous mouse behavior relevant to AD-related behavioral alterations.
May combine two claims.
Interpretationown dataBoth AppNL-G-F and 5xFAD mice exhibit behavioral instability, suggesting that the abnormalities in behavioral structure are an early preclinical feature of AD-related pathogenesis that progresses during clinical stages.
“Thus, both AppNL-G-F and 5xFAD mice exhibit behavioral instability, suggesting that the abnormalities in behavioral structure are an early preclinical feature of AD-related pathogenesis that progresses during clinical stages.”
Quote in the paperSubstantial claimMay combine two claims.
Interpretationown dataAlterations in spontaneous behavior detected by VAME are responsive to AD-related therapeutic interventions and provide unbiased and sensitive outcomes for assessment of treatments.
“They also suggest that alterations in spontaneous behavior detected by VAME are responsive to AD-related therapeutic interventions and provide unbiased and sensitive outcomes for assessment of treatments.”
Quote in the paperSubstantial claimMay combine two claims.
Interpretationown dataFull behavioral sequence analyses revealed significant alterations in the organization of behavioral structure in AppNL-G-F mice, characterized by increased transitions and randomness, and premature escape of fast ambulatory behaviors to engage in slow and static exploration.
“Overall, full behavioral sequence analyses revealed significant alterations in the organization of behavioral structure in AppNL-G-F mice, characterized by increased transitions and randomness, and premature escape of fast ambulatory behaviors to engage in slow and static exploration.”
Quote in the paperSubstantial claim- Rests on an assumption: The 30 postural motifs identified by VAME capture the full repertoire of spontaneous mouse behavior relevant to AD-related behavioral alterations.
May combine two claims.
Interpretationown data5xFAD mice had more pronounced abnormalities in spontaneous behavior than controls and AppNL-G-F mice, indicating a gene-dose-dependent effect.
“5xFAD mice had more pronounced abnormalities in spontaneous behavior than controls and AppNL-G-F mice, indicating a gene-dose-dependent effect.”
Quote in the paperSubstantial claimProbably not a substantial claim: may not pass S3 (load-bearing).
Study design and conductnone givenIn the keypoint-MoSeq analysis, the hyperparameter kappa was configured at 10000 to ensure a syllable length comparable to the output generated by VAME.
“A hyperparameter, kappa, was configured at a value of 10000 to ensure a comparable syllable length to the output generated by VAME.”
Quote in the paperNot a substantial claimProbably not a substantial claim: may not pass S4 (based).
PremisecitationBehavioral or neuropsychiatric changes, such as depression, agitation, anxiety, apathy, irritability, delusions, or sleep disruptions, are often apparent years before cognitive decline or dementia.
“Behavioral or neuropsychiatric changes, such as depression, agitation, anxiety, apathy, irritability, delusions, or sleep disruptions, are often apparent years before cognitive decline or dementia.”
Quote in the paperNot a substantial claimMay combine two claims.
InterpretationreasoningThe organization of spontaneous behavior was largely restored in 5xFAD-Fggγ390-396A mice, indicating that AD-related spontaneous behavioral changes are sensitive to therapeutic interventions and driven by neuroinflammation.
“Notably, the organization of spontaneous behavior was largely restored in 5xFAD-Fggγ390–396A mice, indicating that AD-related spontaneous behavioral changes are sensitive to therapeutic interventions and driven by neuroinflammation.”
Quote in the paperSubstantial claim- Rests on an assumption: The rescue of spontaneous behavioral alterations in 5xFAD-Fggγ390–396A mice is due to reduced AD-related neuroinflammation rather than to other effects of the Fggγ390–396A mutation.
Classified as probably not a substantial claim.
Premisenone givenOur understanding of alterations in spontaneous behavior in AD models has been limited by the technical challenges of assessing non-task-oriented behavior.
“However, our understanding of alterations in spontaneous behavior in AD models has been limited by the technical challenges of assessing non-task-oriented behavior.”
Quote in the paperSubstantial claimProbably not a substantial claim: may not pass S3 (load-bearing).
Study design and conductnone givenThe study does not resolve whether impairments of spontaneous behavior are driven by biological pathways and mechanisms of cognitive decline.
“Another limitation of the study is that it does not resolve whether impairments of spontaneous behavior are driven by biological pathways and mechanisms of cognitive decline.”
Quote in the paperNot a substantial claimProbably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.
Interpretationnone givenOur results and conclusions apply solely to mouse models of AD and our ML pipeline currently has no clinical application.
“It is important to emphasize that our results and conclusions apply solely to mouse models of AD and that our ML pipeline currently has no clinical application.”
Quote in the paperNot a substantial claimProbably not a substantial claim: may not pass S3 (load-bearing).
Study design and conductown dataThe glossary definitions represent the most dominant behavior.
“Glossary definitions represent the most dominant behavior”
Quote in the paperNot a substantial claimProbably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.
InterpretationreasoningML approaches to assess spontaneous behavior in humans can provide early functional outcomes for disease diagnosis and assessment of treatments and have potential to reveal and quantify AD-related behavioral manifestations otherwise difficult to observe.
“However, ML approaches to assess spontaneous behavior in humans, as we have done in mice, can provide early functional outcomes for disease diagnosis and the assessment of treatments and have the potential to reveal and quantify AD-related behavioral manifestations that are otherwise difficult to observe.”
Quote in the paperNot a substantial claimMay combine two claims.
Interpretationown dataThe current approach does not directly assess cognitive functions, but our findings suggest that these outcomes capture key elements of AD pathogenesis, such as inflammation, that could complement traditional task-oriented behavioral approaches.
“The current approach does not directly assess cognitive functions, but our findings suggest that these outcomes capture key elements of AD pathogenesis, such as inflammation, that could complement traditional task-oriented behavioral approaches.”
Quote in the paperSubstantial claimMay combine two claims.
Interpretationown dataVAME segmentation and egocentric coordinates reflect allocentric speeds.
“This finding suggests that VAME segmentation and egocentric coordinates reflect allocentric speeds.”
Quote in the paperSubstantial claimProbably not a substantial claim: may not pass S3 (load-bearing).
Study design and conductown dataThree AppNL-G-F mice were excluded from analysis because they failed to locate the visible platform and had combined latencies and distances for visible training 3 SD above the mean of WT mice.
“Three AppNL–G-F mice were excluded from analysis because they failed to locate the visible platform and had combined latencies and distances for the visible training 3 SD above the mean of WT mice.”
Quote in the paperNot a substantial claimMay combine two claims.
Interpretationown dataSpontaneous behavior was altered in both male and female AppNL-G-F mice, but females were particularly vulnerable to changes in motif usage, motif speed, and randomness of behavioral sequences.
“Spontaneous behavior was altered in both male and female AppNL-G-F mice, but females were particularly vulnerable to changes in motif usage, motif speed, and randomness of behavioral sequences.”
Quote in the paperNot a substantial claimClassified as probably not a substantial claim.
Study design and conductnone givenFor transition network statistics, false discovery rate with Benjamin-Hochberg multiple testing adjustment was used.
“For the transition network statistics, we used the false discovery rate with Benjamin-Hochberg multiple testing adjustment.”
Quote in the paperSubstantial claimMay combine two claims.
InterpretationreasoningAppNL-G-F and 5xFAD mice shared behavioral alterations (e.g., increased motif speed, elevated number of transitions, and reduced dwell time in motifs), suggesting that these deficits are caused by common mechanisms of disease pathogenesis, such as familial AD mutations or abnormally high amyloid levels.
“First, AppNL-G-F and 5xFAD mice shared behavioral alterations (e.g., increased motif speed, elevated number of transitions, and reduced dwell time in motifs), suggesting that these deficits are caused by common mechanisms of disease pathogenesis, such as familial AD mutations or abnormally high amyloid levels.”
Quote in the paperSubstantial claimMay combine two claims.
Interpretationown dataAppNL-G-F mice exhibit robust age-dependent impairments in spontaneous behavior, including motif utilization, disorganized behavioral sequences, increased transitions, and randomness, indicating that these behavioral alterations reflect disease progression.
“We found that AppNL-G-F mice exhibit robust age-dependent impairments in spontaneous behavior, including motif utilization, disorganized behavioral sequences, increased transitions, and randomness, indicating that these behavioral alterations reflect disease progression.”
Quote in the paperSubstantial claim- Rests on an assumption: The 30 postural motifs identified by VAME capture the full repertoire of spontaneous mouse behavior relevant to AD-related behavioral alterations.
Probably not a substantial claim: may not pass S3 (load-bearing).
Study design and conductnone givenFor the side-by-side comparison between VAME and keypoint-MoSeq, independent models were created.
“For the side-by-side comparison between VAME and keypoint-MoSeq, we created independent models.”
Quote in the paperNot a substantial claimClassified as probably not a substantial claim.
Study design and conductnone givenBehavioral experiments were performed in sex-balanced time blocks by investigators unaware of genotype.
“Behavioral experiments were performed in sex-balanced time blocks (morning and afternoon) by investigators who were unaware of the genotype of the mice.”
Quote in the paperSubstantial claimClassified as probably not a substantial claim.
Study design and conductown dataNo held-out group was tested in the classifier analyses.
“no held out group was tested”
Quote in the paperSubstantial claim
The paper relies on this without support.
The 30 postural motifs identified by VAME capture the full repertoire of spontaneous mouse behavior relevant to AD-related behavioral alterations.
“Egocentric coordinates of seven body parts (nose, four paws, center, and tail base) were fed into VAME to train an unsupervised dynamical model, which assessed postural patterns across a sliding window of 16 frames (640-ms window) to identify 30 distinct postural motifs.”
The paper relies on this without support.
The rescue of spontaneous behavioral alterations in 5xFAD-Fggγ390–396A mice is due to reduced AD-related neuroinflammation rather than to other effects of the Fggγ390–396A mutation.
“Fggγ390–396A reduces AD-related pathogenesis, including amyloid burden, dystrophic neurites, synaptic loss, microgliosis, and neuroinflammation in 5xFAD-Fggγ390–396A mice relative to 5xFAD mice.21,22,34”
The paper relies on this without support.
The behavioral abnormalities detected by VAME in mouse models are caused by AD-related brain dysfunction rather than peripheral or motor differences.
“Our findings suggest that ML approaches to assess spontaneous behavior provide a direct, sensitive, and unbiased measure of brain dysfunction induced by AD-related mechanisms in mouse models.”
The paper relies on this without support.
The rescue of behavioral abnormalities in 5xFAD-Fggγ390–396A mice is specifically due to blockade of fibrinogen-microglia interactions.
“Notably, we found that genetic blockade of fibrinogen-microglia interactions in 5xFAD-Fggγ390–396A mice reduced virtually all the motif usage abnormalities found in 5xFAD mice.”
Self-check
Percentages, sums and reported differences re-computed from the paper’s own numbers.
Each contradiction between two passages listed here was confirmed by an independent check.
k-means genotype classification rates conflict with group sizes
MinorWith 14 WT and 18 AppNL-G-F mice, 94.4% cannot be a WT classification rate and 92.8% cannot be an AppNL-G-F classification rate. The reported percentages approximately match 17 of 18 and 13 of 14, respectively, suggesting the genotype labels were reversed.
“k-means model included constant and motifs 17 and 7 (entry order) (correct classification: WT = 94.4% and AppNL–G-F = 92.8%; overall, 93.8%; AIC = 34.749)”
“Middle-age AppNL–G-F cohort: 11–15-month-old AppNL–G-F mice (n = 18; 8 females and 10 males; mean age, 13.25 ± 1.59 [SD] months) and WT littermate controls (n = 14; 8 females and 6 males; 13.27 ± 1.40 months).”
5xFAD classifier omits the 5xFAD class
MinorThe reported three-class classifier names control, Fggγ390-396A, and rescued 5xFAD-Fggγ390-396A mice, but not the unrescued 5xFAD mice. Elsewhere, WT and Fggγ390-396A mice are combined as the control group, so the classifier labels both duplicate a control subgroup and omit a disease group central to the comparison.
“For control, Fggγ390–396A, and 5xFAD-Fggγ390–396A: ML outcomes the model included intercept and motifs 2, 28, 22, 4, and 11 (entry order) (correct classification: control = 100%, Fggγ390–396A = 100%, and 5xFAD-Fggγ390–396A = 100%; AIC = 24.002)”
“Because there were no significant differences in motif usage between Fggγ390–396A and WT mice (Figure S4A), these two groups of mice were combined and treated as control group.”
Cites
Deep Residual Learning for Image Recognition
He, K., Zhang, X., Ren, S., & Sun, J. (2016) · 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr.2016.90 · Cited by 235,474 · see citing works
Fitting Linear Mixed-Effects Models Using <b>lme4</b>
Bates, D., Mächler, M., Bolker, B., & Walker, S. (2015) · Journal of Statistical Software · doi:10.18637/jss.v067.i01 · Cited by 86,835 · see citing works
Cytoscape: A software Environment for integrated models of biomolecular interaction networks
Shannon, P., Markiel, A., Ozier, O., Baliga, N. S., Wang, J. T., Ramage, D., Amin, N., Schwikowski, B., & Ideker, T. (2003) · Genome Research · doi:10.1101/gr.1239303 · Cited by 51,888 · see citing works
McKhann, G. M., Knopman, D. S., Chertkow, H., Hyman, B. T., Jack, C. R., Kawas, C. H., Klunk, W. E., Koroshetz, W. J., Manly, J. J., Mayeux, R., et al. (2011) · Alzheimer's & Dementia · doi:10.1016/j.jalz.2011.03.005 · Cited by 18,657 · cited 1× · see citing works
“10,11 12-14 Thus, these behavioral changes may be the earliest symptoms of preclinical stages of AD.”
Fitting Linear Mixed-Effects Models Using lme4Preprint
Bates, D. M., Maechler, M., Bolker, B. M., & Walker, S. (2014) · arXiv (Cornell University) · doi:10.3929/ethz-b-000105397 · Cited by 9,435 · see citing works
Sperling, R. A., Aisen, P. S., Beckett, L. A., Bennett, D. A., Craft, S., Fagan, A. M., Iwatsubo, T., Jack, C. R., Kaye, J., Montine, T. J., et al. (2011) · Alzheimer's & Dementia · doi:10.1016/j.jalz.2011.03.003 · Cited by 7,155 · cited 1× · see citing works
“10,11 12-14 Thus, these behavioral changes may be the earliest symptoms of preclinical stages of AD.”
DeepLabCut: markerless pose estimation of user-defined body parts with deep learning
Mathis, A., Mamidanna, P., Cury, K. M., Abe, T., Murthy, V. N., Mathis, M. W., & Bethge, M. (2018) · Nature Neuroscience · doi:10.1038/s41593-018-0209-y · Cited by 5,187 · see citing works
Oakley, H., Cole, S. L., Logan, S., Maus, E., Shao, P., Craft, J., Guillozet-Bongaarts, A., Ohno, M., Disterhoft, J., Van Eldik, L., et al. (2006) · The Journal of Neuroscience · doi:10.1523/jneurosci.1202-06.2006 · Cited by 3,530 · cited 1× · see citing works
“16,36,37 Second, spontaneous behavioral abnormalities were age dependent and therefore tracked the progression of AD-related pathological and cognitive alterations.”
The Pathobiology of Vascular Dementia
Iadecola, C. (2013) · Neuron · doi:10.1016/j.neuron.2013.10.008 · Cited by 1,857 · see citing works
Blood–brain barrier breakdown is an early biomarker of human cognitive dysfunction
Nation, D. A., Sweeney, M. D., Montagne, A., Sagare, A. P., D’Orazio, L. M., Pachicano, M., Sepehrband, F., Nelson, A. R., Buennagel, D. P., Harrington, M. G., et al. (2019) · Nature Medicine · doi:10.1038/s41591-018-0297-y · Cited by 1,728 · see citing works
Palop, J. J., Chin, J., Roberson, E. D., Wang, J., Thwin, M. T., Bien-Ly, N., Yoo, J., Ho, K. O., Yu, G.-Q., Kreitzer, A., et al. (2007) · Neuron · doi:10.1016/j.neuron.2007.07.025 · Cited by 1,653 · see citing works
Self-citationUsing DeepLabCut for 3D markerless pose estimation across species and behaviors
Nath, T., Mathis, A., Chen, A. C., Patel, A., Bethge, M., & Mathis, M. W. (2019) · Nature Protocols · doi:10.1038/s41596-019-0176-0 · Cited by 1,476 · cited 1× · see citing works
“…Network analysis of motif transitions reveals increased randomness and altered behavioral sequences in App NL-G-F mice Deconstructing the full sequence of motif-motif transitions may ultimately provide a more sensitive measure of brain function than unitary motif analyses (Figures 2, 3, and 4).”
Single App knock-in mouse models of Alzheimer’s disease
Saito, T., Matsuba, Y., Mihira, N., Takano, J., Nilsson, P., Itohara, S., Iwata, N., & Saido, T. C. (2014) · Nature Neuroscience · doi:10.1038/nn.3697 · Cited by 1,447 · see citing works
Self-citationBayesian Model Choice: Asymptotics and Exact Calculations
Gelfand, A., & Dey, D. (1994) · Journal of the Royal Statistical Society Series B (Statistical Methodology) · doi:10.1111/j.2517-6161.1994.tb01996.x · Cited by 1,399 · cited 1× · see citing works
“: The estimate of the log of pseudo Bayes factors 57 for the Markov chain model comparison between order k and order l is Delta indices for motif transitions, motif use, and motif speed—”
Verret, L., Mann, E. O., Hang, G. B., Barth, A. M. I., Cobos, I., Ho, K., Devidze, N., Masliah, E., Kreitzer, A. C., Mody, I., et al. (2012) · Cell · doi:10.1016/j.cell.2012.02.046 · Cited by 1,216 · cited 1× · see citing works
“16,24,27 Network analysis of motif transitions reveals increased randomness and altered behavioral sequences in App NL-G-F mice Deconstructing the full sequence of motif-motif transitions may ultimately provide a more sensitive measure of brain function than unitary motif analyses (Figures 2, 3,…”
Self-citationIturria-Medina, Y., Sotero, R. C., Toussaint, P. J., Mateos-Pérez, J. M., Evans, A. C., Initiative, T. A. D. N., Weiner, M. W., Aisen, P., Petersen, R., Jack, C. R., et al. (2016) · Nature Communications · doi:10.1038/ncomms11934 · Cited by 1,202 · see citing works
Self-citationSLEAP: A deep learning system for multi-animal pose tracking
Pereira, T., Tabris, N., Matsliah, A., Turner, D. M., Li, J., Ravindranath, S., Papadoyannis, E. S., Normand, E., Deutsch, D. S., Wang, Z. Y., et al. (2022) · Nature Methods · doi:10.1038/s41592-022-01426-1 · Cited by 984 · cited 1× · see citing works
“…Network analysis of motif transitions reveals increased randomness and altered behavioral sequences in App NL-G-F mice Deconstructing the full sequence of motif-motif transitions may ultimately provide a more sensitive measure of brain function than unitary motif analyses (Figures 2, 3, and 4).”
Association of Amyloid and Tau With Cognition in Preclinical Alzheimer Disease: A Longitudinal Study
Hanseeuw, B. J., Betensky, R. A., Jacobs, H. I. L., Schultz, A. P., Sepulcre, J., Becker, J. A., Cosio, D. M. O., Farrell, M., Quiroz, Y. T., Mormino, E. C., et al. (2019) · JAMA Neurology · doi:10.1001/jamaneurol.2019.1424 · Cited by 865 · cited 1× · see citing works
“10,11 12-14 Thus, these behavioral changes may be the earliest symptoms of preclinical stages of AD.”
APP mouse models for Alzheimer’s disease preclinical studies.
Sasaguri, H., Nilsson, P., Hashimoto, S., Nagata, K., Saito, T., De Strooper, B., Hardy, J., Vassar, R., Winblad, B., & Saido, T. C. (2017) · The EMBO Journal · doi:10.15252/embj.201797397 · Cited by 793 · see citing works
Self-citationMapping Sub-Second Structure in Mouse Behavior
Wiltschko, A. B., Johnson, M. J., Iurilli, G., Peterson, R. E., Katon, J. M., Pashkovski, S. L., Abraira, V. E., Adams, R. P., & Datta, S. R. (2015) · Neuron · doi:10.1016/j.neuron.2015.11.031 · Cited by 760 · cited 1× · see citing works
“MoSeq 6 and keypoint-MoSeq 35 use a statistical framework known as autoregressive hidden Markov model (AR-HMM), 6,50,51 whereas VAME 1,7 uses deep recurrent neural networks and variational autoencoders, providing two well-established complementary approaches.”
Fibrinogen in neurological diseases: mechanisms, imaging and therapeutics
Petersen, M. A., Ryu, J. K., & Akassoglou, K. (2018) · Nature reviews. Neuroscience · doi:10.1038/nrn.2018.13 · Cited by 527 · cited 3× · see citing works
“…in 5xFAD mice were prevented by an AD-related therapeutic intervention, Fgg γ 390–396A , that prevents CD11b receptor-mediated interactions with microglia 22,33 and reduces AD-related pathogenesis, including amyloid burden, dystrophic neurites, synaptic loss, microgliosis, and neuroinflammation.”
Self-citationModeling and Prediction of Human Behavior
Pentland, A., & Liu, A. (1999) · Neural Computation · doi:10.1162/089976699300016890 · Cited by 520 · cited 1× · see citing works
“29,30 To assess the predictability of behavior, we fitted Markov chain models at a range of orders (0–3; 640-ms timesteps) (Figure 5B and Table S1).”
Flick, M. J., Du, X., Witte, D. P., Jiroušková, M., Soloviev, D. A., Busuttil, S. J., Plow, E. F., & Degen, J. L. (2004) · Journal of Clinical Investigation · doi:10.1172/jci20741 · Cited by 450 · see citing works
Merlini, M., Rafalski, V. A., Rios Coronado, P. E., Gill, T. M., Ellisman, M., Muthukumar, G., Subramanian, K. S., Ryu, J. K., Syme, C. A., Davalos, D., et al. (2019) · Neuron · doi:10.1016/j.neuron.2019.01.014 · Cited by 431 · cited 5× · see citing works
“21,22,34 Thus, our findings suggest that abnormalities in spontaneous behavior are mechanistically related to AD pathogenesis and may offer mechanistically relevant endpoints for assessing AD-related pathways and therapeutic interventions in mouse models.”
InfluentialSelf-citationThe Striatum Organizes 3D Behavior via Moment-to-Moment Action Selection
Markowitz, J. E., Gillis, W. F., Beron, C. C., Neufeld, S. Q., Robertson, K., Bhagat, N. D., Peterson, R. E., Peterson, E., Hyun, M., Linderman, S. W., et al. (2018) · Cell · doi:10.1016/j.cell.2018.04.019 · Cited by 422 · cited 1× · see citing works
“MoSeq 6 and keypoint-MoSeq 35 use a statistical framework known as autoregressive hidden Markov model (AR-HMM), 6,50,51 whereas VAME 1,7 uses deep recurrent neural networks and variational autoencoders, providing two well-established complementary approaches.”
Ryu, J. K., & McLarnon, J. G. (2009) · Journal of Cellular and Molecular Medicine · doi:10.1111/j.1582-4934.2008.00434.x · Cited by 406 · cited 1× · see citing works
“39-42 Fibrinogen is converted to fibrin, which is deposited at sites of microglia activation in the AD brain.”
Palop, J. J., Jones, B., Kekonius, L., Chin, J., Yu, G.-Q., Raber, J., Masliah, E., & Mucke, L. (2003) · Proceedings of the National Academy of Sciences · doi:10.1073/pnas.1133381100 · Cited by 402 · see citing works
Self-citationThe association of neuropsychiatric symptoms in MCI with incident dementia and Alzheimer disease
Rosenberg, P. B., Mielke, M. M., Appleby, B. S., Oh, E. S., Geda, Y. E., & Lyketsos, C. G. (2013) · The American Journal of Geriatric Psychiatry · doi:10.1016/j.jagp.2013.01.006 · Cited by 366 · cited 5× · see citing works
“These behavioral changes are associated with increased risk of dementia and AD, 13,14 suggesting that they contribute to AD pathogenesis.”
InfluentialCortes-Canteli, M., Paul, J., Norris, E. H., Bronstein, R., Ahn, H. J., Zamolodchikov, D., Bhuvanendran, S., Fenz, K. M., & Strickland, S. (2010) · Neuron · doi:10.1016/j.neuron.2010.05.014 · Cited by 339 · see citing works
GABA-Receptive Microglia Selectively Sculpt Developing Inhibitory Circuits
Favuzzi, E., Huang, S., Saldi, G.-A., Binan, L., Ibrahim, L. A., Fernández-Otero, M., Cao, Y., Zeine, A., Sefah, A., Zheng, K., et al. (2020) · Cell · doi:10.1016/j.cell.2021.06.018 · Cited by 321 · see citing works
Martínez‐Losa, M., Tracy, T. E., Ma, K., Verret, L., Clemente-Perez, A., Khan, A. S., Cobos, I., Ho, K., Gan, L., Mucke, L., et al. (2018) · Neuron · doi:10.1016/j.neuron.2018.02.029 · Cited by 237 · see citing works
Self-citationKeypoint-MoSeq: parsing behavior by linking point tracking to pose dynamics
Weinreb, C., Pearl, J. E., Lin, S., Osman, M. A. M., Zhang, L., Annapragada, S., Conlin, E., Hoffmann, R., Makowska, S., Gillis, W. F., et al. (2024) · Nature Methods · doi:10.1038/s41592-024-02318-2 · Cited by 228 · see citing works
Fibrinogen is associated with an increased risk of Alzheimer disease and vascular dementia
van Oijen, M., Witteman, J. C., Hofman, A., Koudstaal, P. J., & Breteler, M. M. B. (2005) · Stroke · doi:10.1161/01.str.0000189721.31432.26 · Cited by 222 · see citing works
Identifying behavioral structure from deep variational embeddings of animal motion
Luxem, K., Mocellin, P., Fuhrmann, F., Kürsch, J., Miller, S. R., Palop, J. J., Remy, S., & Bauer, P. (2022) · Communications Biology · doi:10.1038/s42003-022-04080-7 · Cited by 218 · cited 6× · see citing works
“MoSeq 6 and keypoint-MoSeq 35 use a statistical framework known as autoregressive hidden Markov model (AR-HMM), 6,50,51 whereas VAME 1,7 uses deep recurrent neural networks and variational autoencoders, providing two well-established complementary approaches.”
InfluentialSelf-citationSpontaneous behaviour is structured by reinforcement without explicit reward
Markowitz, J. E., Gillis, W. F., Jay, M., Wood, J. J., Harris, R., Cieszkowski, R., Scott, R., Brann, D. H., Koveal, D., Kula, T., et al. (2023) · Nature · doi:10.1038/s41586-022-05611-2 · Cited by 214 · cited 1× · see citing works
“MoSeq 6 and keypoint-MoSeq 35 use a statistical framework known as autoregressive hidden Markov model (AR-HMM), 6,50,51 whereas VAME 1,7 uses deep recurrent neural networks and variational autoencoders, providing two well-established complementary approaches.”
Fibrin-targeting immunotherapy protects against neuroinflammation and neurodegeneration
Ryu, J. K., Rafalski, V. A., Meyer-Franke, A., Adams, R. A., Poda, S. B., Rios Coronado, P. E., Pedersen, L. Ø., Menon, V., Baeten, K. M., Sikorski, S. L., et al. (2018) · Nature Immunology · doi:10.1038/s41590-018-0232-x · Cited by 213 · cited 3× · see citing works
“28,34 Fibrin-targeting immunotherapy is currently in phase 1 clinical trials.”
Hultman, K., Strickland, S., & Norris, E. H. (2013) · Journal of Cerebral Blood Flow & Metabolism · doi:10.1038/jcbfm.2013.76 · Cited by 183 · cited 1× · see citing works
“32,43,44 The binding of A β to fibrinogen inhibits fibrin degradation, resulting in chronic fibrin deposition in the AD brain.”
Peters, M. E., Rosenberg, P. B., Steinberg, M., Norton, M. C., Welsh-Bohmer, K. A., Hayden, K. M., Breitner, J., Tschanz, J. T., & Lyketsos, C. G. (2013) · The American Journal of Geriatric Psychiatry · doi:10.1016/j.jagp.2013.01.049 · Cited by 167 · cited 4× · see citing works
“These behavioral changes are associated with increased risk of dementia and AD, 13,14 suggesting that they contribute to AD pathogenesis.”
InfluentialXia, D., Lianoglou, S., Sandmann, T., Calvert, M., Suh, J. H., Thomsen, E., Dugas, J., Pizzo, M. E., DeVos, S. L., Earr, T. K., et al. (2022) · Molecular Neurodegeneration · doi:10.1186/s13024-022-00547-7 · Cited by 124 · cited 4× · see citing works
“16,19 We also tested the role of neuroinflammation in AD-related spontaneous behavioral alterations by blocking fibrinogen-microglia interactions in 5xFAD-Fgg γ 390–396A mice.”
InfluentialSelf-citationDefining blood-induced microglia functions in neurodegeneration through multiomic profiling
Mendiola, A. S., Yan, Z., Dixit, K., Johnson, J. R., Bouhaddou, M., Meyer-Franke, A., Shin, M.-G., Yong, Y., Agrawal, A., MacDonald, E., et al. (2023) · Nature Immunology · doi:10.1038/s41590-023-01522-0 · Cited by 124 · cited 7× · see citing works
“21,22,34 Thus, our findings suggest that abnormalities in spontaneous behavior are mechanistically related to AD pathogenesis and may offer mechanistically relevant endpoints for assessing AD-related pathways and therapeutic interventions in mouse models.”
InfluentialSelf-citationFibrin drives thromboinflammation and neuropathology in COVID-19
Ryu, J. K., Yan, Z., Montano, M., Sozmen, E. G., Dixit, K., Suryawanshi, R. K., Matsui, Y., Helmy, E., Kaushal, P., Makanani, S. K., et al. (2024) · Nature · doi:10.1038/s41586-024-07873-4 · Cited by 116 · cited 2× · see citing works
“28,34 Fibrin-targeting immunotherapy is currently in phase 1 clinical trials.”
Self-citationBiderman, D., Whiteway, M. R., Hurwitz, C., Greenspan, N., Lee, R. S., Vishnubhotla, A., Warren, R., Pedraja, F., Noone, D., Schartner, M. M., et al. (2024) · Nature Methods · doi:10.1038/s41592-024-02319-1 · Cited by 100 · see citing works
Open-source tools for behavioral video analysis: Setup, methods, and best practices
Luxem, K., Sun, J. J., Bradley, S. P., Krishnan, K., Yttri, E., Zimmermann, J., Pereira, T. D., & Laubach, M. (2023) · eLife · doi:10.7554/elife.79305 · Cited by 98 · see citing works
Self-citationBlood–brain barrier leakage and perivascular inflammation in cerebral amyloid angiopathy
Kozberg, M. G., Yi, I., Freeze, W. M., Auger, C. A., Scherlek, A. A., Greenberg, S. M., & van Veluw, S. J. (2022) · Brain Communications · doi:10.1093/braincomms/fcac245 · Cited by 94 · cited 1× · see citing works
“32,43,44 The binding of A β to fibrinogen inhibits fibrin degradation, resulting in chronic fibrin deposition in the AD brain.”
Latif-Hernandez, A., Shah, D., Craessaerts, K., Saido, T., Saito, T., De Strooper, B., Van der Linden, A., & D’Hooge, R. (2019) · Behavioural Brain Research · doi:10.1016/j.bbr.2017.11.017 · Cited by 92 · see citing works
Self-citationXu, G., Zhang, H., Zhang, S., Fan, X., & Liu, X. (2007) · International Journal of Clinical Practice · doi:10.1111/j.1742-1241.2007.01268.x · Cited by 89 · cited 1× · see citing works
“39,45-48 Notably, we found that genetic blockade of fibrinogen-microglia interactions in 5xFAD-Fgg γ 390–396A mice reduced virtually all the motif usage abnormalities found in 5xFAD mice.”
Whyte, L. S., Hemsley, K. M., Lau, A. A., Hassiotis, S., Saito, T., Saido, T. C., Hopwood, J. J., & Sargeant, T. J. (2018) · Behavioural Brain Research · doi:10.1016/j.bbr.2017.09.006 · Cited by 88 · see citing works
Self-citationJohnson, E. C. B., Ho, K., Yu, G.-Q., Das, M., Sanchez, P. E., Djukic, B., Lopez, I., Yu, X., Gill, M., Zhang, W., et al. (2020) · Molecular Neurodegeneration · doi:10.1186/s13024-020-00393-5 · Cited by 73 · cited 2× · see citing works
“16,19 We also tested the role of neuroinflammation in AD-related spontaneous behavioral alterations by blocking fibrinogen-microglia interactions in 5xFAD-Fgg γ 390–396A mice.”
Self-citationPrognostic polypeptide blood plasma biomarkers of Alzheimer's disease progression
Yang, H., Lyutvinskiy, Y., Herukka, S.-K., Soininen, H., Rutishauser, D., & Zubarev, R. A. (2014) · Journal of Alzheimer's Disease · doi:10.3233/jad-132102 · Cited by 52 · cited 1× · see citing works
“39,45-48 Notably, we found that genetic blockade of fibrinogen-microglia interactions in 5xFAD-Fgg γ 390–396A mice reduced virtually all the motif usage abnormalities found in 5xFAD mice.”
Capturing Human Sequence-Learning Abilities in Configuration Design Tasks Through Markov Chains
McComb, C., Cagan, J., & Kotovsky, K. (2017) · Journal of Mechanical Design · doi:10.1115/1.4037185 · Cited by 52 · cited 2× · see citing works
“29,30 To assess the predictability of behavior, we fitted Markov chain models at a range of orders (0–3; 640-ms timesteps) (Figure 5B and Table S1).”
Hidden behavioral fingerprints in epilepsy
Gschwind, T., Zeine, A., Raikov, I., Markowitz, J. E., Gillis, W. F., Felong, S., Isom, L. L., Datta, S. R., & Soltesz, I. (2023) · Neuron · doi:10.1016/j.neuron.2023.02.003 · Cited by 47 · cited 2× · see citing works
“9 MoSeq and VAME are based on unsupervised dimensionality reduction methods, which analyze the implicit structure and patterns of postural motion to identify the most relevant features that capture the spatiotemporal variability of the data to define brief stereotypical postural units.”
Partitioning variability in animal behavioral videos using semi-supervised variational autoencoders
Whiteway, M. R., Biderman, D., Friedman, Y., Dipoppa, M., Buchanan, E. K., Wu, A., Zhou, J., Bonacchi, N., Miska, N. J., Noel, J., et al. (2021) · PLoS Computational Biology · doi:10.1371/journal.pcbi.1009439 · Cited by 44 · see citing works
Sato, K., Watamura, N., Fujioka, R., Mihira, N., Sekiguchi, M., Nagata, K., Ohshima, T., Saito, T., Saido, T. C., & Sasaguri, H. (2021) · Journal of Biological Chemistry · doi:10.1016/j.jbc.2021.101004 · Cited by 37 · cited 1× · see citing works
“16,36,37 Second, spontaneous behavioral abnormalities were age dependent and therefore tracked the progression of AD-related pathological and cognitive alterations.”
Self-citationCho, E., Kim, S., Heo, S.-J., Shin, J., Hwang, S., Kwon, E., Lee, S., Kim, S., & Kang, B. (2023) · Scientific Reports · doi:10.1038/s41598-023-35194-5 · Cited by 32 · see citing works
Self-citationFan, D.-Y., Sun, H.-L., Sun, P.-Y., Jian, J.-M., Li, W.-W., Shen, Y.-Y., Zeng, F., Wang, Y.-J., & Bu, X.-L. (2021) · Frontiers in Neuroscience · doi:10.3389/fnins.2020.625844 · Cited by 27 · see citing works
Fibrin-targeting immunotherapy for dementia
Kantor, A. B., Akassoglou, K., & Stavenhagen, J. B. (2023) · The Journal of Prevention of Alzheimer's Disease · doi:10.14283/jpad.2023.105 · Cited by 24 · cited 1× · see citing works
“49 ML approaches for behavioral segmentation ML approaches for behavioral segmentation have the potential to detect subtle or overlooked behavioral abnormalities linked to early brain dysfunction.”
Self-citationAge-Dependent Behavioral and Metabolic Assessment of AppNL−G−F/NL−G−F Knock-in (KI) Mice
Wang, S., Ichinomiya, T., Savchenko, P., Devulapalli, S., Wang, D., Beltz, G., Saito, T., Saido, T. C., Wagner, S. L., Patel, H. H., et al. (2022) · Frontiers in Molecular Neuroscience · doi:10.3389/fnmol.2022.909989 · Cited by 23 · see citing works
Self-citationA septal-ventral tegmental area circuit drives exploratory behavior
Mocellin, P., Barnstedt, O., Luxem, K., Kaneko, H., Vieweg, S., Henschke, J. U., Dalügge, D., Fuhrmann, F., Karpova, A., Pakan, J. M. P., et al. (2024) · Neuron · doi:10.1016/j.neuron.2023.12.016 · Cited by 16 · cited 1× · see citing works
“…estimation (DeepLabCut, 2,3 SLEAP, 4 and LightningPose 5 ) and behavioral segmentation (MoSeq 6 and variational animal motion embedding [VAME] 7,8 ) make it possible to deconstruct full sequences of spontaneous mouse behavior into brief postural units (motifs or syllables) and reveal their…”
Self-citationLearning Disentangled Behavior Embeddings
Shi, C., Schwartz, S., Levy, S., Achvat, S., Abboud, M., Ghanayim, A., Schiller, J., & Mishne, G. (2021) · neural information processing systems · Cited by 9 · see citing works
Capturing Human Sequence-Learning Abilities in Configuration Design Tasks through Markov ChainsPreprint
McComb, C., Kotovsky, K., & Cagan, J. (2018) · doi:10.31224/osf.io/zetj9 · Cited by 6 · see citing works
Bayesian Model Choice: Asymptotics and Exact Calculations
Gelfand, A. E., & Dey, D. K. (1993) · doi:10.21236/ada269067
Flagged first, then by citations
Cited by
Sorted by citations
Fatima, G., Ashiquzzaman, A., Kim, S. S., Kim, Y. R., Kwon, H.-S., & Chung, E. (2025) · Neurobiology of Disease · doi:10.1016/j.nbd.2025.106877 · Cited by 24 · see citing works
“These cutting-edge approaches enhance the precision, scope, and sensitivity of behavioral analysis, potentially offering valuable insights into disease mechanisms (Miller et al., 2024).”
Zeng, Q., Huang, C., & Zhu, J. (2026) · Frontiers in Public Health · doi:10.3389/fpubh.2025.1656603 · Cited by 20 · see citing works
A decade progress in the phenotyping of App knock-in mouse model of Alzheimer’s disease
Manabe, T., & Saito, T. (2025) · Neuroscience Research · doi:10.1016/j.neures.2025.104959 · Cited by 9 · see citing works
Self-citationDoes advancement in marker-less pose-estimation mean more quality research? A systematic review
Bhola, S., Kim, H., Kim, H. S., Gu, B., & Yoo, J. (2025) · Frontiers in Behavioral Neuroscience · doi:10.3389/fnbeh.2025.1663089 · Cited by 4 · see citing works
Kantor, A. B., Rickert, M., Cheng, H., Flanagan, K., Salgotra, V., Suppahia, A., Ryu, J. K., Widboom, P., Warfield, J. R., Akassoglou, K., et al. (2025) · Journal of Neuroinflammation · doi:10.1186/s12974-025-03650-w · Cited by 5 · see citing works
“Genetic removal of the fibrin inflammatory epitope and treatment with fibrin γ377–395 targeting antibodies restores a favorable homeostatic gene signature and protects from cognitive impairment [20, 21, 30, 33].”
Self-citationLiu, J., Qiao, S., Gong, W., Liu, M., Qifeng, J., Zhang, B., Hui, W., Liu, D., & Zhou, S. Y. (2026) · Chemical Engineering Journal · doi:10.1016/j.cej.2026.173194 · Cited by 3 · see citing works
Afroughi, F., SeyedAlinaghi, S., Mirzapour, P., Shirdel, S., Parmoon, Z., Khorshidi, M. M., Mansouri, S., Sheykhi, M., Popoola, Y., & Mehraeen, E. (2025) · Intelligence-Based Medicine · doi:10.1016/j.ibmed.2025.100338 · Cited by 3 · see citing works
Yan, Z., Mendiola, A. S., Lauderdale, K., Kim, K.-Y., Yong, Y., Leng, K., Bushong, E. A., Schuck, R., Meyer‐Franke, A., Agrawal, A., et al. (2026) · bioRxiv (Cold Spring Harbor Laboratory) · doi:10.64898/2026.05.01.722077 · Cited by 2 · see citing works
Self-citationBoyanova, S., Katsouri, L., Krupic, J., Hair, K., Wang, S.-H., & Wiseman, F. K. (2026) · STAR Protocols · doi:10.1016/j.xpro.2026.104633 · Cited by 1 · see citing works
“168,171 Addressing the limitations of current methods, including lowering the amount of information required for training datasets, improving computational efficiency, developing methods for occluded areas, and improving 3D and group methods, will increase the value of these tools, and thus the…”
Lauderdale, K., Yan, Z., Mendiola, A. S., Zhang, Y., Mallen, D. D., Nambiar, P., Brady, E. S., Miller, S. R., Meza-Acevedo, R., Cabriga, B., et al. (2026) · bioRxiv (Cold Spring Harbor Laboratory) · doi:10.64898/2026.05.01.722324 · Cited by 1 · see citing works
Self-citationSubsecond Analysis of Locomotor Activity in Parkinsonian Mice
Berezhnoi, D., Chehade, H. D., Simms, G., Chen, L., Narasimhan, K., Dravid, S., & Chu, H.-Y. (2024) · eNeuro · doi:10.1523/eneuro.0014-25.2025
Boosting Alzheimer’s Detection using an Ensemble Model on Multi-Feature Subsets
Ashokkumar, P., Vandhana, S., Shankar, G. S., Aravindhan, M., Prabhu, S. N., & Ashok, R. (2025) · doi:10.1109/incsst64791.2025.11210411
Shin, J., Brady, E. S., Chen, C., Lauderdale, K., Agrawal, A., Zhang, Y., Jiang, X., Nambiar, P., Herbert, J., Mallen, D. D., et al. (2025) · bioRxiv (Cold Spring Harbor Laboratory) · doi:10.1101/2025.11.11.687887
“To determine early signs of dysfunction induced by E4 and A b , we applied the ML-VAME platform to segment naturally occurring spontaneous behavior into brief behavioral “motifs” as mice freely explored an open arena 40,41 .”
Self-citationTisdale, R. K., Sun, Y., Miller, S. R., Lee, S. M., Park, S., Shin, J., Allocca, G., Palop, J. J., & Kilduff, T. S. (2025) · bioRxiv (Cold Spring Harbor Laboratory) · doi:10.64898/2025.12.29.696818
“Figure 4 presents the time in each state, activity and subcutaneous body temperature (T sc ) based on 24-h recordings collected from WT and App NL-G-F mice at 14-18 months of age and again at 18-22 months.”
Self-citationAI-Augmented Computational Modeling of Human Behavior
Zhang, G., Zhang, S., & Wu, H. (2025) · Springer eBooks · doi:10.1007/978-981-97-8440-0_111-1
Sleep-Wake Transitions Are Impaired in the AppNL-G-F Mouse Model of Early Onset Alzheimer’s DiseasePreprint
Tisdale, R. K., Sun, Y., Miller, S. R., Lee, S. M., Park, S., Shin, J., Allocca, G., Palop, J. J., & Kilduff, T. S. (2026) · eLife · doi:10.7554/elife.110240.1
Self-citationTisdale, R. K., Sun, Y., Miller, S. R., Lee, S. M., Park, S., Shin, J., Allocca, G., Palop, J. J., & Kilduff, T. S. (2026) · eLife · doi:10.7554/elife.110240
Self-citationDavis, M. N., Bullock, M., Jhaldiyal, A., Holifield, N., Mikhail, K., Voskobiynyk, Y., Vollmer, R. M., Prendergast, G. C., Lauderdale, K., Palop, J. J., et al. (2026) · eNeuro · doi:10.1523/eneuro.0304-25.2026
“New computer-vision and machine-learning approaches provide a more sensitive and unbiased tool for behavioral analysis (Nath et al., 2019; Miller et al., 2024; Weinreb et al., 2024).”
Self-citationAI-Augmented Computational Modeling of Human Behavior
Zhang, G., Zhang, S., & Wu, H. (2026) · Springer eBooks · doi:10.1007/978-981-95-3658-0_111
Tisdale, R. K., Sun, Y., Miller, S. R., Lee, S. M., Park, S., Shin, J., Allocca, G., Palop, J. J., & Kilduff, T. S. (2026) · eLife · doi:10.7554/elife.110240.2
Self-citationKashyap, S. N., Fox, S. N., DeNittis, V. R., Wilson, K. I., Boyle, N. R., Bullock, M., Tadepalli, A. S., Miller, S. R., Palop, J. J., Ambaw, Y. A., et al. (2026) · Science Translational Medicine · doi:10.1126/scitranslmed.adr9158 · Cited by 3 · see citing works
Self-citationCognitive Drift Mapping Using Ambient Micro-Behavioral IoT Signals in Preclinical Neurodegeneration
Aarthi, S., & Ravikumar, R. N. (2026) · Advances in computational intelligence and robotics book series · doi:10.4018/979-8-2600-0462-3.ch012
Cid-Galiot, J. J., Lasserre, A. A. A., Cuervo-Zanatta, D., eu-repo, J. E. D. H. I., & Trujillo-Mata, A. (2026) · Biosciences Biotechnology Research Asia · doi:10.13005/bbra/3547
- Authors
- Miller, S. R., Luxem, K., Lauderdale, K., Nambiar, P., Honma, P. S., Ly, K. K., Bangera, S., Bullock, M., Shin, J., Kaliss, N., Qiu, Y., Cai, C., Shen, K., Mallen, K. D., Yan, Z., Mendiola, A. S., Saito, T., Saido, T. C., Pico, A. R., Thomas, R., Roberson, E. D., Akassoglou, K., Bauer, P., Remy, S., & Palop, J. J.
- Affiliations
- University of California, San Francisco (US)
- German Center for Neurodegenerative Diseases (DE)
- Gladstone Institutes (US)
- University of Alabama at Birmingham (US)
- Nagoya City University (JP)
- RIKEN Center for Brain Science (JP)
- Corresponding author
- Miller, Stephanie R.
- Palop, Jorge J.
- ORCID
- Venue
- Cell Reports
- Publisher
- Elsevier BV
- Published
- 2024-11
- Volume
- 43
- Issue
- 11
- Pages
- 114870
- ISSN
- 2211-1247 (ISSN-L)
- PMC
- PMC12010505
- OpenAlex
- W4403565324
- Semantic Scholar
- 89845ed64b389418c96a0cd2f1b08f29f4fce894
- Cited by
- 24
- References
- 60
- Citations by year
- 2026: 13 · 2025: 10 · 2024: 1
- FWCI
- 6.83
- Type
- journal article
- Language
- en
- Open access
- gold
- doi.org
- Licence
- Locations
- Cell Reports (publishedVersion, cc-by)
- PubMed (publishedVersion)
- PubMed Central (acceptedVersion, cc-by)
- DZNE Pub (submittedVersion, cc-by-nc-nd)
- DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion, cc-by-sa)
- Informationszentrum Lebenswissenschaften (ZB MED) (submittedVersion, cc-by-nc-nd)
- Funders
- California Department of Fish and Game (RTG 2413 SynAGE)
- National Institute of Neurological Disorders and Stroke (R35NS097976, K99 NS126707)
- National Institutes of Health (AG082147)
- National Institute on Aging (RF1AG06926, R01AG061785, RF1AG062234, K01-AG083732, R01AG073082, RF1AG059405, P01AG062629)
- Gladstone Institutes
- European Research Council
- Topics
- Health Sciences › Medicine › Physiology › Alzheimer's disease research and treatments
- Life Sciences › Neuroscience › Neurology › Neuroinflammation and Neurodegeneration Mechanisms
- Health Sciences › Medicine › Psychiatry and Mental health › Dementia and Cognitive Impairment Research
- Keywords
- Alzheimer's disease, spontaneous behavior, behavioral phenotyping, humanized mice, transgenic mouse model, knockin mouse model, treatment efficacy, neuroinflammation, machine learning
- MeSH terms
- Machine Learning, Alzheimer Disease, Animals, Behavior, Animal, Disease Models, Animal, Female, Humans, Male, Mice, Inbred C57BL, Mice, Transgenic, Amyloid beta-Protein Precursor, Treatment Outcome, Mice
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
Claims read by deepseek-v4.1-flash under definition version 3 (extract-v3) on 2026-10-30.
Classified by jev-1.13-20260917 (gold-guideline-v0.2) on 2026-10-30.
Cited sources are checked by gpt-6-sol.
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
