CoChat

Functional alterations of the prefrontal circuit underlying cognitive aging in mice

Open access
Chong, Huee Ru, Ranjbar-Slamloo, Yadollah, Ho, Malcolm Zheng Hao, Ouyang, Xuan, Kamigaki, Tsukasa · 2023 · Nature Communications

The DOI matches its Crossref record; the references it cites are not checked yet.

Abstract

Executive function is susceptible to aging. How aging impacts the circuit-level computations underlying executive function remains unclear. Using calcium imaging and optogenetic manipulation during memory-guided behavior, we show that working-memory coding and the relevant recurrent connectivity in the mouse medial prefrontal cortex (mPFC) are altered as early as middle age. Population activity in the young adult mPFC exhibits dissociable yet overlapping patterns between tactile and auditory modalities, enabling crossmodal memory coding concurrent with modality-dependent coding. In middle age, however, crossmodal coding remarkably diminishes while modality-dependent coding persists, and both types of coding decay in advanced age. Resting-state functional connectivity, especially among memory-coding neurons, decreases already in middle age, suggesting deteriorated recurrent circuits for memory maintenance. Optogenetic inactivation reveals that the middle-aged mPFC exhibits heightened vulnerability to perturbations. These findings elucidate functional alterations of the prefrontal circuit that unfold in middle age and deteriorate further as a hallmark of cognitive aging.

Overview

The claims and statistics of this paper haven’t been checked yet.

Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.

A few free checks a day · papers already verified open instantly · results are public

Integrity signals

Registry

Crossref · 10.1038/s41467-023-43142-0

Cites

Statistics

Figures

Fig. 1 Aging effects on behavioral performance of the bimodal delayed 2-AFC task. a Schematic of the task design. Mouse cartoons modified from the drawing73. b, c Number of daily sessions to reach a behavioral criterion (b) and correct rate during the imaging sessions (c) (n = 11 young mice, n = 7 middle-aged mice, n = 5 advanced-aged mice). Open circles, individual mice. *U = 81, p = 0.032, **U = 72.5, p = 0.015, in (b), *U = 59, p = 0.030 in (c), Wilcoxon rank-sum test. Box plots show the median and the 25th and 75th percentiles as box edges and the whiskers extend to the 5th and 95th percentiles. Y young; M Middle age, A advanced age. Source data are provided as a Source Data file.
Fig. 1 Aging effects on behavioral performance of the bimodal delayed 2-AFC task. a Schematic of the task design. Mouse cartoons modified from the drawing73. b, c Number of daily sessions to reach a behavioral criterion (b) and correct rate during the imaging sessions (c) (n = 11 young mice, n = 7 middle-aged mice, n = 5 advanced-aged mice). Open circles, individual mice. *U = 81, p = 0.032, **U = 72.5, p = 0.015, in (b), *U = 59, p = 0.030 in (c), Wilcoxon rank-sum test. Box plots show the median and the 25th and 75th percentiles as box edges and the whiskers extend to the 5th and 95th percentiles. Y young; M Middle age, A advanced age. Source data are provided as a Source Data file.CC BY · PubMed Central
Fig. 2 Aging effects on action-plan coding in the mPFC during the bimodal d2-AFC task. a Calcium imaging with the GRIN lens. Left, modified from the 3D Brain Explorer Allen Reference Atlas74, atlas.brain-map.org. Right, modified from Allen Mouse Brain Coronal section (image 38)74, atlas.brain-map.org. b Example field of view (maximum intensity projection of dF/F movie). Scale bar, 100 μm. c Activity trace of 10 example ROIs extracted from (b). Blue and red vertical lines, sample onset in Left and Right trials, respectively. Gray shading, delay period. d Trial-averaged activity (z-scored) of the example action-plan-selective delay neurons from each age group. Gray shading, delay period; vertical line, sample onset. Left column, Left-delay neurons; right column, Right-delay neurons. Colored shadings, ± s.e.m. e Fraction of action-plan-selective delay neurons for the imaging sessions (young, n = 23 sessions, 11 mice; middle age, n = 14 sessions, 7 mice; advanced age, n = 10 sessions, 5 mice). Left, early delay period, *U = 504, p = 0.037, **U = 485, p = 2.5 × 10−4; Right, late delay period, *U = 505, p = 0.035, **U = 472, p = 1.6 × 10−3, Wilcoxon rank-sum test. Box plots show the median and the 25th and 75th percentiles as box edges and the whiskers extend to the 5th and 95th percentiles. Y, young; M, middle age; A, advanced age. f Mean absolute selectivity index for action plan ( | SI | ) (young, n = 2,245 neurons; middle age, n = 859 neurons; advanced age, n = 968 neurons). Error bars, ± s.e.m. **p < 0.001, one-way ANOVA with Bonferroni post hoc test. g Decoding accuracy of action plan in each task period. Decoding accuracy during the early and late delay (gray shading) in the middle-aged and the advanced-aged groups was significantly lower than that in the young group (p < 0.001, two-way ANOVA with Bonferroni post hoc test). Error bars, ± s.e.m. across 100 iterations. Horizontal dotted lines, 97.5th and 2.5th percentiles of the shuffled data. Bsl baseline, Smp sample, eDly early delay, lDly late delay, Fbc feedback period. Source data are provided as a Source Data file.
Fig. 2 Aging effects on action-plan coding in the mPFC during the bimodal d2-AFC task. a Calcium imaging with the GRIN lens. Left, modified from the 3D Brain Explorer Allen Reference Atlas74, atlas.brain-map.org. Right, modified from Allen Mouse Brain Coronal section (image 38)74, atlas.brain-map.org. b Example field of view (maximum intensity projection of dF/F movie). Scale bar, 100 μm. c Activity trace of 10 example ROIs extracted from (b). Blue and red vertical lines, sample onset in Left and Right trials, respectively. Gray shading, delay period. d Trial-averaged activity (z-scored) of the example action-plan-selective delay neurons from each age group. Gray shading, delay period; vertical line, sample onset. Left column, Left-delay neurons; right column, Right-delay neurons. Colored shadings, ± s.e.m. e Fraction of action-plan-selective delay neurons for the imaging sessions (young, n = 23 sessions, 11 mice; middle age, n = 14 sessions, 7 mice; advanced age, n = 10 sessions, 5 mice). Left, early delay period, *U = 504, p = 0.037, **U = 485, p = 2.5 × 10−4; Right, late delay period, *U = 505, p = 0.035, **U = 472, p = 1.6 × 10−3, Wilcoxon rank-sum test. Box plots show the median and the 25th and 75th percentiles as box edges and the whiskers extend to the 5th and 95th percentiles. Y, young; M, middle age; A, advanced age. f Mean absolute selectivity index for action plan ( | SI | ) (young, n = 2,245 neurons; middle age, n = 859 neurons; advanced age, n = 968 neurons). Error bars, ± s.e.m. **p < 0.001, one-way ANOVA with Bonferroni post hoc test. g Decoding accuracy of action plan in each task period. Decoding accuracy during the early and late delay (gray shading) in the middle-aged and the advanced-aged groups was significantly lower than that in the young group (p < 0.001, two-way ANOVA with Bonferroni post hoc test). Error bars, ± s.e.m. across 100 iterations. Horizontal dotted lines, 97.5th and 2.5th percentiles of the shuffled data. Bsl baseline, Smp sample, eDly early delay, lDly late delay, Fbc feedback period. Source data are provided as a Source Data file.CC BY · PubMed Central
Fig. 3 The unimodal delayed 2-AFC task and behavioral performance. a,Sequence of the unimodal d2-AFC task. Tactile and auditory trial blocks were alternated with the bimodal trial block, and each block was switched every 20–30 trials. Mouse cartoons modified from the drawing73. b Correct rate on the first day tested with the unimodal d2-AFC task as well as the bimodal task for all the mice (n = 10 young mice, n = 7 middle-aged mice, n = 5 advanced-aged mice). *U = 60, p = 0.02, **U = 101, p = 0.008, Wilcoxon rank-sum test. Box plots show the median and the 25th and 75th percentiles as box edges and the whiskers extend to the 5th and 95th percentiles. Y young, M middle age, A advanced age. Source data are provided as a Source Data file.
Fig. 3 The unimodal delayed 2-AFC task and behavioral performance. a,Sequence of the unimodal d2-AFC task. Tactile and auditory trial blocks were alternated with the bimodal trial block, and each block was switched every 20–30 trials. Mouse cartoons modified from the drawing73. b Correct rate on the first day tested with the unimodal d2-AFC task as well as the bimodal task for all the mice (n = 10 young mice, n = 7 middle-aged mice, n = 5 advanced-aged mice). *U = 60, p = 0.02, **U = 101, p = 0.008, Wilcoxon rank-sum test. Box plots show the median and the 25th and 75th percentiles as box edges and the whiskers extend to the 5th and 95th percentiles. Y young, M middle age, A advanced age. Source data are provided as a Source Data file.CC BY · PubMed Central
Fig. 4 Modality-general coding degrades with aging. a Trial-averaged activity of the example consistent action-plan-selective delay neurons in the unimodal task. Gray shading, delay; vertical line, sample onset. Colored shadings, ± s.e.m. b Correlation of selectivity index (SI) between tactile and auditory trials for the imaging sessions (young, n = 19 sessions, 10 mice; middle age, n = 14 sessions, 7 mice; advanced age, n = 10 sessions, 5 mice). Box plots, the median and the 25th and 75th percentiles as box edges with the whiskers extending to the 5th and 95th percentiles. *U = 401, p = 0.0048, **U = 359, p = 7.5 × 10−4 for early delay, *U = 397, p = 0.0074, **U = 362, p = 4.5 × 10−4 for late delay, Wilcoxon rank-sum test. Y, young; M, middle age; A, advanced age. c, SI in tactile × SI in auditory trials averaged across all neurons (young, n = 1,577 neurons; middle age, n = 689 neurons; advanced age, n = 606 neurons). Error bars, ± s.e.m. **p < 0.001, one-way ANOVA with Bonferroni post hoc test. d Fraction of consistent, tactile, and auditory action-plan-selective delay neurons (Methods) (young, n = 19 sessions, 10 mice; middle age, n = 14 sessions, 7 mice; advanced age, n = 10 sessions, 5 mice). Error bars, ± s.e.m. e Example trajectories of population activity in state space (only 3 PC axes shown for display purposes, Methods). Filled circle, delay onset; open circle, delay end. f Similarity index in state space. **p < 0.001, one-way ANOVA with Bonferroni post hoc test. Error bars, ± s.d. across 1000 iterations. g Within- and cross-modality decoding of action plan (Methods). Horizontal lines, 97.5th and 2.5th percentile of the shuffled distribution. Error bars, ± s.e.m. across 100 iterations. h Product of multiplying within-modality and cross-modality decoding accuracy. **p < 0.001, one-way ANOVA with Bonferroni post hoc test. Error bars, ± s.e.m. across 100 iterations. i Modality-action-plan decoding (Methods). **p < 0.001, one-way ANOVA with Bonferroni post hoc test. Error bars, ± s.e.m. across 100 iterations. Source data are provided as a Source Data file.
Fig. 4 Modality-general coding degrades with aging. a Trial-averaged activity of the example consistent action-plan-selective delay neurons in the unimodal task. Gray shading, delay; vertical line, sample onset. Colored shadings, ± s.e.m. b Correlation of selectivity index (SI) between tactile and auditory trials for the imaging sessions (young, n = 19 sessions, 10 mice; middle age, n = 14 sessions, 7 mice; advanced age, n = 10 sessions, 5 mice). Box plots, the median and the 25th and 75th percentiles as box edges with the whiskers extending to the 5th and 95th percentiles. *U = 401, p = 0.0048, **U = 359, p = 7.5 × 10−4 for early delay, *U = 397, p = 0.0074, **U = 362, p = 4.5 × 10−4 for late delay, Wilcoxon rank-sum test. Y, young; M, middle age; A, advanced age. c, SI in tactile × SI in auditory trials averaged across all neurons (young, n = 1,577 neurons; middle age, n = 689 neurons; advanced age, n = 606 neurons). Error bars, ± s.e.m. **p < 0.001, one-way ANOVA with Bonferroni post hoc test. d Fraction of consistent, tactile, and auditory action-plan-selective delay neurons (Methods) (young, n = 19 sessions, 10 mice; middle age, n = 14 sessions, 7 mice; advanced age, n = 10 sessions, 5 mice). Error bars, ± s.e.m. e Example trajectories of population activity in state space (only 3 PC axes shown for display purposes, Methods). Filled circle, delay onset; open circle, delay end. f Similarity index in state space. **p < 0.001, one-way ANOVA with Bonferroni post hoc test. Error bars, ± s.d. across 1000 iterations. g Within- and cross-modality decoding of action plan (Methods). Horizontal lines, 97.5th and 2.5th percentile of the shuffled distribution. Error bars, ± s.e.m. across 100 iterations. h Product of multiplying within-modality and cross-modality decoding accuracy. **p < 0.001, one-way ANOVA with Bonferroni post hoc test. Error bars, ± s.e.m. across 100 iterations. i Modality-action-plan decoding (Methods). **p < 0.001, one-way ANOVA with Bonferroni post hoc test. Error bars, ± s.e.m. across 100 iterations. Source data are provided as a Source Data file.CC BY · PubMed Central
Fig. 5 Resting-state functional connectivity (RSFC) degrades with aging. a Schematic of the functional connectivity of mPFC neurons. b Top, RSFC as a function of the inter-neuronal distance. Task-relevant RSFC (blue lines) was computed between pairs of action-plan-selective delay neurons that had the same sign of action-plan selectivity for each imaging session (young, n = 17,427 pairs from 19 sessions; middle-age, n = 4506 pairs from 14 sessions; advanced-age, n = 5277 pairs from 10 sessions). Non-task-relevant RSFC (light-blue lines) was computed between action-plan-selective delay neurons and non-task-relevant neurons (young, n = 23,787 pairs from 20 sessions; middle age, n = 8175 pairs from 14 sessions; advanced age, n = 17,093 pairs from 10 sessions). Error bars denote ± s.e.m. Inset, the average calcium event rate during the resting state was not different between action-plan-selective delay neurons (blue) and non-task-relevant neurons (gray) for any of the age groups (U = 379, p = 0.82 for young, U = 197, p = 0.80 for middle age, U = 117, p = 0.39 for advanced age; Wilcoxon rank-sum test). Young, n = 1072 action-plan-selective delay neurons and 394 non-task-relevant neurons, 19 sessions; middle age, n = 408 action-plan-selective delay neurons and 258 non-task-relevant neurons, 14 sessions; advanced age, n = 444 action-plan-selective delay neurons and 417 non-task-relevant neurons, 10 sessions. Bottom, same as top, but after shuffling the calcium events across the time bins. Blue and light-blue lines, median of the shuffled distribution for the task-relevant and non-task-relevant RSFC, respectively. Dashed lines, 2.5th and 97.5th percentile. Note the smaller scale compared with the top panels. c Difference between the task-relevant and non-task-relevant RSFC. Open circles, each inter-neuronal distance bin. *p < 0.02, **p < 0.0005, one-way ANOVA with Bonferroni post hoc test. Error bars denote ± s.e.m. Y, young; M, middle age; A, advanced age. Source data are provided as a Source Data file.
Fig. 5 Resting-state functional connectivity (RSFC) degrades with aging. a Schematic of the functional connectivity of mPFC neurons. b Top, RSFC as a function of the inter-neuronal distance. Task-relevant RSFC (blue lines) was computed between pairs of action-plan-selective delay neurons that had the same sign of action-plan selectivity for each imaging session (young, n = 17,427 pairs from 19 sessions; middle-age, n = 4506 pairs from 14 sessions; advanced-age, n = 5277 pairs from 10 sessions). Non-task-relevant RSFC (light-blue lines) was computed between action-plan-selective delay neurons and non-task-relevant neurons (young, n = 23,787 pairs from 20 sessions; middle age, n = 8175 pairs from 14 sessions; advanced age, n = 17,093 pairs from 10 sessions). Error bars denote ± s.e.m. Inset, the average calcium event rate during the resting state was not different between action-plan-selective delay neurons (blue) and non-task-relevant neurons (gray) for any of the age groups (U = 379, p = 0.82 for young, U = 197, p = 0.80 for middle age, U = 117, p = 0.39 for advanced age; Wilcoxon rank-sum test). Young, n = 1072 action-plan-selective delay neurons and 394 non-task-relevant neurons, 19 sessions; middle age, n = 408 action-plan-selective delay neurons and 258 non-task-relevant neurons, 14 sessions; advanced age, n = 444 action-plan-selective delay neurons and 417 non-task-relevant neurons, 10 sessions. Bottom, same as top, but after shuffling the calcium events across the time bins. Blue and light-blue lines, median of the shuffled distribution for the task-relevant and non-task-relevant RSFC, respectively. Dashed lines, 2.5th and 97.5th percentile. Note the smaller scale compared with the top panels. c Difference between the task-relevant and non-task-relevant RSFC. Open circles, each inter-neuronal distance bin. *p < 0.02, **p < 0.0005, one-way ANOVA with Bonferroni post hoc test. Error bars denote ± s.e.m. Y, young; M, middle age; A, advanced age. Source data are provided as a Source Data file.CC BY · PubMed Central
Fig. 6 Optogenetic inactivation of the mPFC reveals age-dependent vulnerability to perturbations. a Optogenetic inactivation during the whole delay period. Pyramidal neurons in the mPFC expressing stGtACR2 were inactivated with blue light. b Left, unilateral inactivation. The correct rate decreased both in young (young, n = 6 mice × 2 hemispheres, black lines) and middle-aged mice (middle age, n = 6 mice × 2 hemispheres, magenta lines) (p < 0.0001 both for young and middle-aged, one-sided bootstrap, Methods). Right, bilateral inactivation. The correct rate decreased both in young (n = 6 mice) and middle-aged mice (n = 6 mice) (p < 0.0001 both for young and middle-aged mice). Faint colored lines, individual mice. c Behavioral effect size measured as the difference in the correct rate between laser-off and laser-on conditions. Bilateral inactivation produced a larger effect size than unilateral inactivation in both age groups (p < 0.0001 both for young and middle-aged mice, one-sided bootstrap). The effect size was not different between the two age groups in either unilateral (p = 0.11, one-sided bootstrap) or bilateral inactivation (p = 0.29, one-sided bootstrap). d Optogenetic inactivation during the early delay period. e Both unilateral and bilateral inactivation decreased the correct rate both in young (black lines) and middle-aged mice (magenta lines) (p < 0.0001 for all comparisons). f Behavioral effect size was significantly larger in the middle-aged mice than the young mice both in unilateral (p < 0.0001, one-sided bootstrap) and bilateral inactivation (p < 0.0001, one-sided bootstrap). Bilateral inactivation produced a larger effect size than unilateral inactivation in both age groups (p < 0.0001 for both of the age groups, one-sided bootstrap). Mouse brain cartoons (b, c, e, f, bottom) modified from the drawing73. Y, young; M, middle age. Source data are provided as a Source Data file.
Fig. 6 Optogenetic inactivation of the mPFC reveals age-dependent vulnerability to perturbations. a Optogenetic inactivation during the whole delay period. Pyramidal neurons in the mPFC expressing stGtACR2 were inactivated with blue light. b Left, unilateral inactivation. The correct rate decreased both in young (young, n = 6 mice × 2 hemispheres, black lines) and middle-aged mice (middle age, n = 6 mice × 2 hemispheres, magenta lines) (p < 0.0001 both for young and middle-aged, one-sided bootstrap, Methods). Right, bilateral inactivation. The correct rate decreased both in young (n = 6 mice) and middle-aged mice (n = 6 mice) (p < 0.0001 both for young and middle-aged mice). Faint colored lines, individual mice. c Behavioral effect size measured as the difference in the correct rate between laser-off and laser-on conditions. Bilateral inactivation produced a larger effect size than unilateral inactivation in both age groups (p < 0.0001 both for young and middle-aged mice, one-sided bootstrap). The effect size was not different between the two age groups in either unilateral (p = 0.11, one-sided bootstrap) or bilateral inactivation (p = 0.29, one-sided bootstrap). d Optogenetic inactivation during the early delay period. e Both unilateral and bilateral inactivation decreased the correct rate both in young (black lines) and middle-aged mice (magenta lines) (p < 0.0001 for all comparisons). f Behavioral effect size was significantly larger in the middle-aged mice than the young mice both in unilateral (p < 0.0001, one-sided bootstrap) and bilateral inactivation (p < 0.0001, one-sided bootstrap). Bilateral inactivation produced a larger effect size than unilateral inactivation in both age groups (p < 0.0001 for both of the age groups, one-sided bootstrap). Mouse brain cartoons (b, c, e, f, bottom) modified from the drawing73. Y, young; M, middle age. Source data are provided as a Source Data file.CC BY · PubMed Central

Verification

Claims the paper makes

The claims and statistics of this paper haven’t been checked yet.

Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.

A few free checks a day · papers already verified open instantly · results are public

Claims other papers make about this one

Each is checked against this paper’s own text: does it say what the citing paper says it does?

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

The self-check needs the paper’s full text.

Citations

The citation lists are queued

They are fetched later and appear here then.

References 72 · Cited by 19

Research works printed but not linked: 1

  1. Flurkey, K., Currer, J. M. & Harrison, D. E. In The Mouse in Biomedical Research, 2nd Edition. (eds Fox, J. G et al.) Vol. 3, 637–672 (New York, Elsevier, 2007). · DOI · Paper page

Metadata

Authors
Chong, H. R., Ranjbar-Slamloo, Y., Ho, M. Z. H., Ouyang, X., & Kamigaki, T.
Affiliations
Nanyang Technological University (SG)
Corresponding author
Kamigaki, Tsukasa
Venue
Nature Communications
Published
2023-11-09
Volume
14
Issue
1
Pages
7254-7254
ISSN
2041-1723 (ISSN-L)
OpenAlex
W4388522989
Cited by
19
References
72
Citations by year
2026: 5 · 2025: 9 · 2024: 4 · 2023: 1
FWCI
2.84
Type
journal article
Language
en
Open access
Licence
Locations
  • Nature Communications (publishedVersion, cc-by)
  • PubMed Central (publishedVersion, cc-by)
  • PubMed (publishedVersion)
  • RePEc: Research Papers in Economics (submittedVersion)
  • DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion, cc-by-sa)
  • DR-NTU (Nanyang Technological University) (submittedVersion, cc-by)
Funders
Ministry of Education - Singapore (RS01/19, RG19/21, LKC MOE, MOE-T2EP30220-0003)
Topics
  • Life Sciences › Neuroscience › Cognitive Neuroscience › Neural dynamics and brain function
  • Life Sciences › Neuroscience › Cognitive Neuroscience › Memory and Neural Mechanisms
  • Life Sciences › Neuroscience › Cellular and Molecular Neuroscience › Neuroscience and Neuropharmacology Research
Keywords
cognitive aging, medial prefrontal cortex, executive function, working memory, recurrent connectivity, resting-state functional connectivity, memory-guided behavior, optogenetic inactivation, middle age (human life stage)
MeSH terms
Cognitive Aging, Aging, Animals, Magnetic Resonance Imaging, Memory, Short-Term, Neural Pathways, Prefrontal Cortex, 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.

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

How to cite

Cite CoChat for the checks on this page, and the paper for its own findings.

This page
CoChat. (2026). Citation check: Functional alterations of the prefrontal circuit underlying cognitive aging in mice. https://app.cochat.ai/papers/10.1038/s41467-023-43142-0
The paper
Chong, H. R., Ranjbar-Slamloo, Y., Ho, M. Z. H., Ouyang, X., & Kamigaki, T. (2023). Functional alterations of the prefrontal circuit underlying cognitive aging in mice. Nature Communications. https://doi.org/10.1038/s41467-023-43142-0

Data sources

Bibliographic data: Semantic Scholar, OpenAlex, Crossref.

Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).

Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.