The DOI matches its Crossref record; 160 of the 160 references it cites were found in a registry or an index; 51 of the 63 sources its claims cite were checked for support.
Transfer learning enables predictions in network biology
Open accessAbstract
Mapping gene networks requires large amounts of transcriptomic data to learn the connections between genes, which impedes discoveries in settings with limited data, including rare diseases and diseases affecting clinically inaccessible tissues. Recently, transfer learning has revolutionized fields such as natural language understanding 1,2 and computer vision 3 by leveraging deep learning models pretrained on large-scale general datasets that can then be fine-tuned towards a vast array of downstream tasks with limited task-specific data. Here, we developed a context-aware, attention-based deep learning model, Geneformer, pretrained on a large-scale corpus of about 30 million single-cell transcriptomes to enable context-specific predictions in settings with limited data in network biology. During pretraining, Geneformer gained a fundamental understanding of network dynamics, encoding network hierarchy in the attention weights of the model in a completely self-supervised manner. Fine-tuning towards a diverse panel of downstream tasks relevant to chromatin and network dynamics using limited task-specific data demonstrated that Geneformer consistently boosted predictive accuracy. Applied to disease modelling with limited patient data, Geneformer identified candidate therapeutic targets for cardiomyopathy. Overall, Geneformer represents a pretrained deep learning model from which fine-tuning towards a broad range of downstream applications can be pursued to accelerate discovery of key network regulators and candidate therapeutic targets.
Overview
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
53 Own result · 23 Own interpretation · 20 Supported · 14 Partly supported · 9 No basis given · 6 Couldn’t confirm · 6 Own method · 1 Supported so far · 1 No full text of the sources · accuracy not yet measured
The cited sources that could be checked couldn’t confirm or dispute this claim.
In the paper
“Furthermore, genes known to be highly context-dependent, such as NOTCH receptors, showed more variability in their embeddings across variable cell types14 compared to the known housekeeping gene GAPDH (Extended Data Fig. 3).”
§ Transfer learning enables predictions in network biology > Context-awareness and batch integration
What the sources say
- Could not verifycites [14] Li, Y. et al. Single-cell transcriptome analysis reveals dynamic cell populations and differential gene expression patterns in control and aneurysmal human aortic tissue. Circulation 142, 1374–1388 (2020). sourceExperimental — accuracy not yet measured
The Results sections describe cell-type-specific gene expression in human aortic tissue, but the passages found do not discuss NOTCH receptors as context-dependent genes. Searches for NOTCH, GAPDH, and embedding variability found no passage addressing the cited point. The embedding comparison is attributed to the citing paper’s Extended Data Fig. 3, not clearly to this source. (LLM, conf=0.72)
plaincorrelational- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Could not verify
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“Mapping the gene regulatory networks that drive disease progression enables screening for molecules that correct the network by normalizing core regulatory elements, rather than targeting peripheral downstream effectors that may not be disease modifying4,5.”
§ Transfer learning enables predictions in network biology > Abstract
What the sources say
- Partly supportedcites [4] Theodoris, C. V. et al. Human disease modeling reveals integrated transcriptional and epigenetic mechanisms of NOTCH1 haploinsufficiency. Cell 160, 1072–1086 (2015). sourceExperimental — accuracy not yet measured
Split into 2 atoms — 1 confirmed, 1 partially supported, 0 disputed, 0 misattributed, 0 unverifiable. Expand the claim to see per-atom verdicts.
- Supportedcites [5] Theodoris, C. V. et al. Network-based screen in iPSC-derived cells reveals therapeutic candidate for heart valve disease. Science 371, eabd0724 (2021). source
Split into 2 atoms — 2 confirmed, 0 partially supported, 0 disputed, 0 misattributed, 0 unverifiable. Expand the claim to see per-atom verdicts.
plaindirect- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Partly supported
Every source the paper cites for this claim was checked and supports it.
In the paper
“Using a publicly available aortic aneurysm dataset14 as a test case, we found that while the original data was impacted by inter-patient variability, Geneformer cell embeddings clustered primarily by cell type and phenotype as opposed to individual patient (Extended Data Fig. 4a).”
§ Transfer learning enables predictions in network biology > Context-awareness and batch integration
What the sources say
- Supportedcites [14] Li, Y. et al. Single-cell transcriptome analysis reveals dynamic cell populations and differential gene expression patterns in control and aneurysmal human aortic tissue. Circulation 142, 1374–1388 (2020). source
The citation identifies the aortic aneurysm dataset used as a test case; the Geneformer clustering result is presented as the citing authors’ own finding. The source paper’s Methods describe single-cell RNA sequencing of aortic tissue from eight patients with ascending thoracic aortic aneurysm and three controls, and state that the data will be made available to other researchers. It therefore supports the dataset identification, without needing to report the Geneformer result. (LLM, conf=0.92)
plaincorrelational- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Supported
Every source the paper cites for this claim was checked and supports it.
In the paper
“Embeddings of genes in iPSC-derived myogenic cells16 showed similar context awareness with in silico differentiation via MYOD (Extended Data Fig. 2d–e).”
§ Transfer learning enables predictions in network biology > Context-awareness and batch integration
What the sources say
- Supportedcites [16] Guo, D. et al. iMyoblasts for ex vivo and in vivo investigations of human myogenesis and disease modeling. eLife 11, e70341 (2022). source
The citation is attached to “iPSC-derived myogenic cells,” identifying the cells used, rather than to the citing authors’ embedding result. The Results section (“Isolation of iMyoblasts by iPSC transgene-free induction and reserve cell selection”) describes generating myogenic iMyoblasts from iPSCs. The source need not report the embedding analysis to support this identification. (LLM, conf=0.91)
plaincorrelational- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Supported
Every source the paper cites for this claim was checked and supports it.
In the paper
“When we in silico reprogrammed fibroblasts15 by artificially adding OCT4, SOX2, KLF4, and MYC to the front of their rank value encodings, the remaining genes in the transcriptome significantly shifted their embedding towards the iPSC state (Extended Data Fig. 2b–c).”
§ Transfer learning enables predictions in network biology > Context-awareness and batch integration
What the sources say
- Supportedcites [15] Xing, Q. R. et al. Diversification of reprogramming trajectories revealed by parallel single-cell transcriptome and chromatin accessibility sequencing. Sci. Adv. 6, 463–474 (2020). source
The citation is attached to “fibroblasts,” identifying the source of the cells or data used in the citing authors’ in silico analysis. The source paper’s Materials and Methods (Experimental design) states that reprogramming began with BJ fibroblasts and that their cells were profiled by single-cell RNA sequencing. The embedding shift is presented as the citing authors’ own result, not a finding attributed to this source. (LLM, conf=0.90)
plaindirect- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Supported
Every source the paper cites for this claim was checked and supports it.
In the paper
“In silico deletion of GATA4 had a significantly higher effect on genes known to be most significantly dysregulated by GATA4 variants in a previously reported iPSC disease model of GATA4-related heart defects33 (Extended Data Fig. 9e).”
§ Transfer learning enables predictions in network biology > In silico gene network analysis
What the sources say
- Supportedcites [33] Ang, Y. S. et al. Disease model of GATA4 mutation reveals transcription factor cooperativity in human cardiogenesis. Cell 167, 1734–1749 (2016). source
The Results section, “Attenuated Cardiac Gene Program in Mutant CPCs and CMs” (page unavailable), reports an iPSC-based GATA4 G296S disease model and identifies genes differentially expressed in mutant cells. Citation 33 identifies the previously reported model and its dysregulated genes; the in silico deletion result is the citing paper’s own finding, not a result attributed to this source. (LLM, conf=0.91)
plaindirect- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Supported
Every source the paper cites for this claim was checked and supports it.
In the paper
“Notably, direct GATA4 targets (as defined by ChIP-seq33) were significantly more impacted by in silico deletion of GATA4 in fetal cardiomyocytes compared to indirect targets (Fig. 5a).”
§ Transfer learning enables predictions in network biology > In silico gene network analysis
What the sources say
- Supportedcites [33] Ang, Y. S. et al. Disease model of GATA4 mutation reveals transcription factor cooperativity in human cardiogenesis. Cell 167, 1734–1749 (2016). source
The citation is attached to “as defined by ChIP-seq,” not to the in silico deletion result, which the citing sentence assigns to its own Figure 5a. In Results (“Genome-Wide Co-Occupancy of GATA4 and TBX5 in Human CMs”; page unavailable), the source reports using ChIP-seq to identify GATA4 binding and validate direct targets. (LLM, conf=0.91)
plaindirect- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Supported
Every source the paper cites for this claim was checked and supports it.
In the paper
“Indeed, in silico deletion of GATA4 or TBX5 alone had a significantly more deleterious effect on their known co-bound targets33 compared to housekeeping genes.”
§ Transfer learning enables predictions in network biology > In silico gene network analysis
What the sources say
- Supportedcites [33] Ang, Y. S. et al. Disease model of GATA4 mutation reveals transcription factor cooperativity in human cardiogenesis. Cell 167, 1734–1749 (2016). source
The citation is attached to “known co-bound targets.” In Results (Genome-Wide Co-Occupancy of GATA4 and TBX5 in Human CMs), the source identifies gene targets co-bound by GATA4 and TBX5. The in silico deletion comparison is the citing authors’ finding, not the point this citation identifies. (LLM, conf=0.92)
plaindirect- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Supported
Every source the paper cites for this claim was checked and supports it.
In the paper
“When examining the attention weights in aortic ECs14, we found that 20% of attention heads significantly attended transcription factors more than other genes, indicating that specific attention heads learned, in an entirely self-supervised manner, the relative importance of transcription factors in distinguishing cell states (Fig. 4e).”
§ Transfer learning enables predictions in network biology > Pretraining encoded network hierarchy
What the sources say
- Supportedcites [14] Li, Y. et al. Single-cell transcriptome analysis reveals dynamic cell populations and differential gene expression patterns in control and aneurysmal human aortic tissue. Circulation 142, 1374–1388 (2020). source
The citation is attached to “aortic ECs,” identifying the cells examined, rather than to the citing authors’ 20% attention-head finding. In Results, under “Fibroblasts and Other Nonimmune Cells in Human Ascending Aortic Wall” (page unavailable), the source identifies two clusters of endothelial cells in human aortic tissue. The 20% result is presented as the citing authors’ own analysis. (LLM, conf=0.91)
plaincorrelational- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Supported
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“Analogously, in silico deletion of TBX5, another known congenital heart disease gene, in fetal cardiomyocytes23 more significantly impacted its direct targets (as defined by ChIP-seq34) compared to indirect targets and housekeeping genes (Extended Data Fig. 9f).”
§ Transfer learning enables predictions in network biology > In silico gene network analysis
What the sources say
- Supportedcites [23] Cao, J. et al. A human cell atlas of fetal gene expression. Science 370, 808 (2020). source
The Discussion (page unavailable) describes a human fetal single-cell gene-expression atlas, and the Identification and annotation section (page unavailable) says the profiled organs include the heart. Citation 23 follows “fetal cardiomyocytes” and identifies the source of the fetal-cell data; the citing sentence does not attribute the TBX5 deletion result to this atlas. (LLM, conf=0.82)
- Partly supportedcites [34] Kathiriya, I. S. et al. Modeling human TBX5 haploinsufficiency predicts regulatory networks for congenital heart disease. Dev. Cell 56, 292–309 (2021). sourceExperimental — accuracy not yet measured
The Results sections on cardiomyocyte function and transcriptional responses report TBX5 loss-of-function in iPSC-derived cardiomyocytes and an association between TBX5-sensitive genes and TBX5 binding sites. They do not report an in silico deletion in fetal cardiomyocytes. The direct-versus-indirect-target comparison is pointed to the citing paper’s Extended Data Fig. 9f, rather than clearly attributed to this source. (LLM, conf=0.85)
plaindirect- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Supported
- Probably not a substantial claim. 23
- It may state two separate claims as one. 27
- The quote only partly states this claim. 3
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“The advent of the self-attention mechanism1,2 has further transformed the deep learning field by generating context-aware models that are able to pay attention to large input spaces and learn which elements are most important to focus on in each context, boosting predictions in a wide realm of applications2,8.”
§ Transfer learning enables predictions in network biology > Abstract
What the sources say
- Partly supportedcites [1] Vaswani, A. et al. Attention is all you need. Preprint at https://doi.org/10.48550/arXiv.1706.03762 (2017). sourceExperimental — accuracy not yet measured
The Background (p. 2) describes self-attention as relating positions within a sequence and reports its successful use in several tasks. Section 3.2.3 (p. 5) says each encoder position can attend to all positions in the preceding layer, and the Machine Translation results (p. 8) report improved translation scores. These passages support the mechanism and some performance gains, but do not establish that it transformed deep learning or boosted predictions across a wide realm of applications. (LLM, conf=0.84)
- Not checked — no full text of the sourcecites [2] Devlin, J., Chang, M. W., Lee, K. & Toutanova, K. BERT: pre-training of deep bidirectional transformers for language understanding. In Proc. 2019 Conference North American Chapter of the Association for Computational Linguistics: Human Language Technologies Vol. 1 (eds Burstein, J. et al.) 4174–4186 (Association for Computational Linguistics, 2019). source
- Supportedcites [8] Lin, T., Wang, Y., Liu, X. & Qiu, X. A survey of transformers. Preprint at https://doi.org/10.48550/arXiv.2106.04554 (2021). source
Section 2.4.1 describes self-attention as using weights generated from relationships between inputs, allowing it to model long-range dependencies and handle variable-length inputs. Section 4.1 says each token can attend to all other tokens, while the Introduction and Section 8 report strong performance across a wide range of applications. These passages support the citing sentence’s broad paraphrase, though Section 4 notes that long inputs pose a computational challenge. (LLM, conf=0.87)
plaindirect- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: fails — The paper’s conclusions may not depend on it.
- Based: passes — The paper gives a reason why it holds.
Probably not a substantial claim. The paper’s conclusions may not depend on it. It may state two separate claims as one.
- Partly supported
Every source the paper cites for this claim was checked and supports it.
In the paper
“Interestingly, the fine-tuned Geneformer’s cell embeddings primarily clustered by cell types and showed improved integration of platforms compared to the original data even after batch effect removal using the ComBat17 or Harmony18 methods (Extended Data Fig. 4b–f).”
§ Transfer learning enables predictions in network biology > Context-awareness and batch integration
What the sources say
- Supportedcites [17] Zhang, Y., Parmigiani, G. & Johnson, W. E. ComBat-seq: batch effect adjustment for RNA-seq count data. NAR Genom. Bioinform. 2, lqaa078 (2020). source
The citation is attached to the ComBat method used for batch-effect removal, not to the citing authors’ Geneformer results. The Discussion identifies ComBat as a batch-effect adjustment approach, and Materials and Methods describes applying it to logged, normalized data. (LLM, conf=0.90)
- Supportedcites [18] Korsunsky, I. et al. Fast, sensitive and accurate integration of single-cell data with Harmony. Nat. Methods 16, 1289–1296 (2019). source
The citation to Korsunsky et al. identifies Harmony as a batch-effect removal method; it does not attribute the Geneformer comparison to that paper. The Online Methods section describes Harmony as producing a batch-corrected cell embedding. (LLM, conf=0.93)
plaincorrelational- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
It may state two separate claims as one.
- Supported
The cited sources that could be checked couldn’t confirm or dispute this claim.
In the paper
“Recently, transfer learning has revolutionized fields such as natural language understanding1,2 and computer vision3 by leveraging deep learning models pretrained on large-scale general datasets that can then be fine-tuned towards a vast array of downstream tasks with limited task-specific data.”
§ Transfer learning enables predictions in network biology > Abstract
What the sources say
- Could not verifycites [1] Vaswani, A. et al. Attention is all you need. Preprint at https://doi.org/10.48550/arXiv.1706.03762 (2017). sourceExperimental — accuracy not yet measured
Split into 2 atoms — 0 confirmed, 0 partially supported, 0 disputed, 0 misattributed, 2 unverifiable. Expand the claim to see per-atom verdicts.
- Not checked — no full text of the sourcecites [2] Devlin, J., Chang, M. W., Lee, K. & Toutanova, K. BERT: pre-training of deep bidirectional transformers for language understanding. In Proc. 2019 Conference North American Chapter of the Association for Computational Linguistics: Human Language Technologies Vol. 1 (eds Burstein, J. et al.) 4174–4186 (Association for Computational Linguistics, 2019). source
- Not checked — no full text of the sourcecites [3] He, K., Zhang, X., Ren, S. & Sun, J. Deep residual learning for image recognition. In Proc. IEEE Computer Society Conference on Computer Vision and Pattern Recognition 770–778 (IEEE, 2016). source
plaindirect- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: fails — The paper’s conclusions may not depend on it.
- Based: passes — The paper gives a reason why it holds.
Probably not a substantial claim. The paper’s conclusions may not depend on it.
- Could not verify
The cited sources that could be checked couldn’t confirm or dispute this claim.
In the paper
“We found that the gene embeddings were robust to sequencing platform11, preservation method12,13, and individual patient variability14 (Extended Data Fig. 2a).”
§ Transfer learning enables predictions in network biology > Context-awareness and batch integration
What the sources say
- Could not verifycites [11] Selewa, A. et al. Systematic comparison of high-throughput single-cell and single-nucleus transcriptomes during cardiomyocyte differentiation. Sci. Rep. 10, 1535 (2020). sourceExperimental — accuracy not yet measured
The Methods and Results compare Drop-seq and DroNc-seq transcriptomes, but no passage assesses Geneformer gene embeddings or their robustness to sequencing platform. The Results instead report substantial differences in gene detection between the two techniques; that does not establish whether the citing authors’ embeddings were robust. Citation 11 is attached only to the sequencing-platform part of the sentence. (LLM, conf=0.78)
- Not matchedcites [12] 10x Genomics Datasets https://www.10xgenomics.com/resources/datasets/frozen-pbm-cs-donor-a-1-standard-1-1-0.
Couldn’t be matched to a paper in Crossref, so it can’t be checked.
- Not matchedcites [13] 10X Genomics Datasets https://www.10xgenomics.com/resources/datasets/fresh-68-k-pbm-cs-donor-a-1-standard-1-1-0.
Couldn’t be matched to a paper in Crossref, so it can’t be checked.
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
It may state two separate claims as one.
- Could not verify
The cited sources that could be checked couldn’t confirm or dispute this claim.
In the paper
“Pretraining was accomplished using a masked learning objective, which has been shown in other informational fields1,2 to improve generalizability of the foundational knowledge learned during pretraining for a wide range of downstream fine-tuning objectives.”
§ Transfer learning enables predictions in network biology > Geneformer architecture and pretraining
What the sources say
- Could not verifycites [1] Vaswani, A. et al. Attention is all you need. Preprint at https://doi.org/10.48550/arXiv.1706.03762 (2017). sourceExperimental — accuracy not yet measured
The Background (p. 2) discusses self-attention across several tasks, and §6.3 (p. 10) reports Transformer results on constituency parsing. Neither passage addresses whether a masked pretraining objective improves generalizability for downstream fine-tuning; searches for masked learning and pretraining found no passage that does. (LLM, conf=0.74)
- Not checked — no full text of the sourcecites [2] Devlin, J., Chang, M. W., Lee, K. & Toutanova, K. BERT: pre-training of deep bidirectional transformers for language understanding. In Proc. 2019 Conference North American Chapter of the Association for Computational Linguistics: Human Language Technologies Vol. 1 (eds Burstein, J. et al.) 4174–4186 (Association for Computational Linguistics, 2019). source
boosteddirect- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
Probably not a substantial claim.
- Could not verify
The sources with a verdict support this claim; the rest weren’t checked yet, have no full text we could read, or their verdicts aren’t rated yet.
In the paper
“The model identified 478 genes whose loss was predicted to shift cardiomyocytes towards dilated cardiomyopathy, which were enriched for pathways involved in muscle contraction38 and mitochondrial39 function.”
§ Transfer learning enables predictions in network biology > In silico treatment analysis
What the sources say
- Not checked — no full text of the sourcecites [38] Kamisago, M. et al. Mutations in sarcomere protein genes as a cause of dilated cardiomyopathy. New Engl. J. Med. 343, 1688–1696 (2000). source
- Supportedcites [39] Ramaccini, D. et al. Mitochondrial function and dysfunction in dilated cardiomyopathy. Front. Cell Dev. Biol. https://doi.org/10.3389/fcell.2020.624216 (2021). source
Citation 39 is attached to mitochondrial function, not to the citing authors’ model or its count of 478 genes. The source’s “Mitochondrial-Associated Dilated Cardiomyopathy” section (page unavailable) describes genetic defects in mitochondrial pathways associated with cardiomyopathy, supporting the cited connection between mitochondrial function and dilated cardiomyopathy. (LLM, conf=0.88)
plaindirect- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
It may state two separate claims as one.
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“Again, Geneformer significantly boosted the ability to predict central versus peripheral factors compared to alternative methods (AUC 0.81) (Fig. 4a, Extended Data Fig. 8e).”
§ Transfer learning enables predictions in network biology > Network dynamics predictions
What the sources say
- Partly supportedcites [32] Litviňuková, M. et al. Cells of the adult human heart. Nature 588, 455–472 (2020). sourceExperimental — accuracy not yet measured
Split into 2 atoms — 0 confirmed, 1 partially supported, 0 disputed, 0 misattributed, 1 unverifiable. Expand the claim to see per-atom verdicts.
boosteddirect- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
The quote only partly states this claim.
- Partly supported
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“Overall, the model identified 447 genes whose loss was predicted to shift cardiomyocytes towards the hypertrophic cardiomyopathy state, which were enriched for pathways including Titin binding36 and sarcomere organization37 known to impact hypertrophic cardiomyopathy pathogenesis.”
§ Transfer learning enables predictions in network biology > In silico treatment analysis
What the sources say
- Partly supportedcites [36] Hinson, J. T. et al. Titin mutations in iPS cells define sarcomere insufficiency as a cause of dilated cardiomyopathy. Science 349, 982–986 (2015). sourceExperimental — accuracy not yet measured
Split into 2 atoms — 0 confirmed, 2 partially supported, 0 disputed, 0 misattributed, 0 unverifiable. Expand the claim to see per-atom verdicts.
- Supportedcites [37] Seidman, C. E. & Seidman, J. G. Identifying sarcomere gene mutations in hypertrophic cardiomyopathy: a personal history. Circ. Res. 108, 743–750 (2011). source
Split into 2 atoms — 2 confirmed, 0 partially supported, 0 disputed, 0 misattributed, 0 unverifiable. Expand the claim to see per-atom verdicts.
plaindirect- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
It may state two separate claims as one.
- Partly supported
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“Top enriched pathways for hypertrophic cardiomyopathy pointed to candidate cardiomyocyte-specific therapeutic targets including ADCY5, disruption of which is associated with longevity and protection against cardiomyopathy in mouse models40, as well as druggable targets41 including SRPK3, a downstream effector of MEF242 which is known to play a critical role in myocardial cell hypertrophy43.”
§ Transfer learning enables predictions in network biology > In silico treatment analysis
What the sources say
- Supportedcites [40] Ho, D., Yan, L., Iwatsubo, K., Vatner, D. E. & Vatner, S. F. Modulation of β-adrenergic receptor signaling in heart failure and longevity: targeting adenylyl cyclase type 5. Heart Fail. Rev. 15, 495–512 (2010). source
The citation is attached to the claim about ADCY5 disruption, not to the citing authors’ pathway-enrichment result or the statements about SRPK3. In “Longevity in mice with AC5 disruption” (page unavailable), the source reports that AC5-knockout mice live longer and are protected from aging-induced cardiomyopathy. (LLM, conf=0.94)
- Partly supportedcites [41] Wagner, A. H. et al. DGIdb 2.0: mining clinically relevant drug-gene interactions. Nucleic Acids Res. 44, D1036–D1044 (2016). sourceExperimental — accuracy not yet measured
The Introduction (page unavailable) describes DGIdb as a resource for checking whether candidate genes belong to potentially druggable categories, supporting the citation’s use for identifying druggable targets. Our copy does not identify SRPK3 as druggable or link it to hypertrophic cardiomyopathy, so it does not support that specific example. (LLM, conf=0.83)
- Supportedcites [42] Nakagawa, O. et al. Centronuclear myopathy in mice lacking a novel muscle-specific protein kinase transcriptionally regulated by MEF2. Genes Dev. 19, 2066–2077 (2005). source
Citation 42 is attached to the description of SRPK3 as a downstream effector of MEF2, not to the pathway-enrichment or druggability claims. The Discussion identifies Srpk3 as a muscle-specific protein kinase directly regulated by MEF2. (LLM, conf=0.93)
plaincorrelational- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
It may state two separate claims as one.
- Supported
Every source the paper cites for this claim was checked and supports it.
In the paper
“We previously mapped the NOTCH1 (N1)-dependent gene network governing cardiac valve disease and identified central regulatory nodes whose correction had broad restorative impact on the network at large4,5.”
§ Transfer learning enables predictions in network biology > Network dynamics predictions
What the sources say
- Supportedcites [4] Theodoris, C. V. et al. Human disease modeling reveals integrated transcriptional and epigenetic mechanisms of NOTCH1 haploinsufficiency. Cell 160, 1072–1086 (2015). source
The Results section describes a predicted NOTCH1-dependent gene network and identifies SOX7 and TCF4 as highly connected regulatory nodes. It reports that correcting those nodes restored expression of the majority of a 48-gene panel toward the wild-type state in patient-derived endothelial cells; the Discussion describes the effect as reaching a large portion of the dysregulated network. (LLM, conf=0.93)
- Supportedcites [5] Theodoris, C. V. et al. Network-based screen in iPSC-derived cells reveals therapeutic candidate for heart valve disease. Science 371, eabd0724 (2021). source
The “Map of gene network dysregulated by N1 haploinsufficiency” section describes mapping the N1-regulated network and identifying SOX7 and TCF4 as highly connected nodes. The “Network-correcting molecules” section reports that targeting those nodes broadly corrected dysregulated genes. Page numbers are not provided in our copy. (LLM, conf=0.94)
plaindirect- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
It may state two separate claims as one.
- Supported
Every source the paper cites for this claim was checked and supports it.
In the paper
“TEAD4 is a transcription factor involved in the Hippo signaling pathway27, and future work is warranted to further examine its role in cardiac development.”
§ Transfer learning enables predictions in network biology > Gene dosage sensitivity predictions
What the sources say
- Supportedcites [27] Currey, L., Thor, S. & Piper, M. TEAD family transcription factors in development and disease. Development 148, dev196675 (2021). source
In “TEADs in vertebrates,” Fig. 1 describes TEAD1-4 as binding YAP/TAZ to activate transcription when Hippo signalling is off. This supports TEAD4’s involvement in the Hippo pathway; the sentence’s call for future work on cardiac development is not attributed to this source. (LLM, conf=0.94)
plaincorrelational- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: fails — The paper’s conclusions may not depend on it.
- Based: passes — The paper gives a reason why it holds.
Probably not a substantial claim. The paper’s conclusions may not depend on it.
- Supported
Every source the paper cites for this claim was checked and supports it.
In the paper
“Additionally, fine-tuning the pretrained Geneformer on the Heart Atlas ECs32 was able to distinguish N1 downstream targets from non-targets without any perturbation data, further demonstrating the model’s ability to encode key features of gene network dynamics and again significantly boosting predictions compared to alternative methods (Fig. 4b, Extended Data Fig. 9a).”
§ Transfer learning enables predictions in network biology > Network dynamics predictions
What the sources say
- Supportedcites [32] Litviňuková, M. et al. Cells of the adult human heart. Nature 588, 455–472 (2020). source
The citation is attached to “Heart Atlas ECs,” identifying the data used for fine-tuning rather than sourcing the Geneformer performance claim. The source’s “Vascular, stromal and mesothelial cells” section describes ten endothelial-cell populations in its adult human heart atlas (page not available in our copy). (LLM, conf=0.92)
boosteddirect- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
It may state two separate claims as one.
- Supported
Every source the paper cites for this claim was checked and supports it.
In the paper
“Among the top 25 deleted genes with the most significant effect were transcription factors known to regulate myocardial development (e.g. FOXM125,26) and entirely novel dosage-sensitive gene candidates such as TEAD4 (Supplementary Table 3).”
§ Transfer learning enables predictions in network biology > Gene dosage sensitivity predictions
What the sources say
- Supportedcites [25] Bolte, C. et al. Expression of Foxm1 transcription factor in cardiomyocytes is required for myocardial development. PLoS ONE 6, e22217 (2011). source
The citation is attached to FOXM1 as an example of a transcription factor known to regulate myocardial development, not to the citing paper’s top-25 ranking or TEAD4 finding. The Introduction reports that deleting Foxm1 from cardiomyocytes disrupts myocardial development and states that Foxm1 is required for proper heart development. Page numbers were not supplied for these passages. (LLM, conf=0.93)
- Supportedcites [26] Bolte, C. et al. Postnatal ablation of Foxm1 from cardiomyocytes causes late onset cardiac hypertrophy and fibrosis without exacerbating pressure overload-induced cardiac remodeling. PLoS ONE 7, e48713 (2012). source
The citation is attached to FOXM1, so the relevant attributed claim is that FOXM1 is a transcription factor known to regulate myocardial development. The Introduction states that Foxm1 is a transcription factor and plays an essential role in embryonic cardiac development. The top-25 ranking and TEAD4 claim are not attributed to this source. (LLM, conf=0.93)
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
It may state two separate claims as one.
- Supported
Every source the paper cites for this claim was checked and supports it.
In the paper
“Geneformer fine-tuned with only Drop-seq data automatically integrated DroNc-seq data such that the fine-tuned Geneformer cell embeddings primarily clustered by cell types and showed improved integration of platforms compared to the original data even after batch effect removal using the ComBat17 or Harmony18 methods.”
§ Transfer learning enables predictions in network biology > Extended Data
What the sources say
- Supportedcites [17] Zhang, Y., Parmigiani, G. & Johnson, W. E. ComBat-seq: batch effect adjustment for RNA-seq count data. NAR Genom. Bioinform. 2, lqaa078 (2020). source
Split into 2 atoms — 2 confirmed, 0 partially supported, 0 disputed, 0 misattributed, 0 unverifiable. Expand the claim to see per-atom verdicts.
- Supportedcites [18] Korsunsky, I. et al. Fast, sensitive and accurate integration of single-cell data with Harmony. Nat. Methods 16, 1289–1296 (2019). source
Split into 2 atoms — 2 confirmed, 0 partially supported, 0 disputed, 0 misattributed, 0 unverifiable. Expand the claim to see per-atom verdicts.
plaindirect- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
It may state two separate claims as one.
- Supported
The cited sources that could be checked couldn’t confirm or dispute this claim.
In the paper
“Pretraining was accomplished using a masked learning objective, which has been shown in other informational fields1,2 to improve generalizability of the foundational knowledge learned during pretraining for a wide range of downstream fine-tuning objectives.”
§ Transfer learning enables predictions in network biology > Methods > Geneformer architecture and pretraining
What the sources say
- Could not verifycites [1] Vaswani, A. et al. Attention is all you need. Preprint at https://doi.org/10.48550/arXiv.1706.03762 (2017). sourceExperimental — accuracy not yet measured
Section 3.2.3 (p. 5) describes masking decoder attention to preserve autoregressive prediction, not a masked pretraining objective. The Abstract (p. 1) and Section 6.3 (pp. 9–10) discuss applying the Transformer to translation and parsing, but do not show that masked pretraining improves generalizability across downstream fine-tuning objectives. Searched for masked learning, pretraining, and fine-tuning. (LLM, conf=0.76)
- Not checked — no full text of the sourcecites [2] Devlin, J., Chang, M. W., Lee, K. & Toutanova, K. BERT: pre-training of deep bidirectional transformers for language understanding. In Proc. 2019 Conference North American Chapter of the Association for Computational Linguistics: Human Language Technologies Vol. 1 (eds Burstein, J. et al.) 4174–4186 (Association for Computational Linguistics, 2019). source
plaindirect- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
Probably not a substantial claim.
- Could not verify
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“consistent with observations that large-scale pretraining allows training of deeper models that ultimately have greater predictive potential in fields including NLU, computer vision, and mathematical problem-solving44.”
§ Transfer learning enables predictions in network biology > Discussion
What the sources say
- Partly supportedcites [44] Henighan, T. et al. Scaling laws for autoregressive generative modeling. Preprint at https://doi.org/10.48550/arXiv.2010.14701 (2020). sourceExperimental — accuracy not yet measured
Section 3.4 (p. 15) reports that larger pretrained image models achieve better classification performance than models trained from scratch. Section 2.1 (p. 6) says image and math tasks favor deeper models than language, and Appendix D (p. 31) reports improved language question-answering with model size. The paper does not establish that large-scale pretraining enables training deeper models across all three fields; it distinguishes model size, depth, and the benefit of pretraining. (LLM, conf=0.85)
plaindirect- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: fails — The paper’s conclusions may not depend on it.
- Based: passes — The paper gives a reason why it holds.
Probably not a substantial claim. The paper’s conclusions may not depend on it.
- Partly supported
Every source the paper cites for this claim was checked and supports it.
In the paper
“Effect on cardiomyocyte embeddings from in silico deletion of genes linked by prior transcriptome-wide association study (TWAS)-prioritized GWAS24 to cardiac MRI traits relevant to cardiac pathology (left ventricular (LV) end diastolic volume (EDV), LV end systolic volume (LVESV), LV ejection fraction (LVEF), and stroke volume (SV)) compared to in silico deletion of control cardiac disease genes”
§ Transfer learning enables predictions in network biology > Extended Data
What the sources say
- Supportedcites [24] Pirruccello, J. P. et al. Analysis of cardiac magnetic resonance imaging in 36,000 individuals yields genetic insights into dilated cardiomyopathy. Nat. Commun. 11, 2254 (2020). source
Split into 2 atoms — 2 confirmed, 0 partially supported, 0 disputed, 0 misattributed, 0 unverifiable. Expand the claim to see per-atom verdicts.
plaindirect- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
The quote only partly states this claim.
- Supported
The cited sources that could be checked couldn’t confirm or dispute this claim.
In the paper
“We found that the gene embeddings were robust to sequencing platformM11, preservation method13,12, and individual patient variability14.”
§ Transfer learning enables predictions in network biology > Extended Data
What the sources say
- Not matchedcites M11
Couldn’t be matched to a paper in Crossref, so it can’t be checked.
- Not matchedcites [13] 10X Genomics Datasets https://www.10xgenomics.com/resources/datasets/fresh-68-k-pbm-cs-donor-a-1-standard-1-1-0.
Couldn’t be matched to a paper in Crossref, so it can’t be checked.
- Not matchedcites [12] 10x Genomics Datasets https://www.10xgenomics.com/resources/datasets/frozen-pbm-cs-donor-a-1-standard-1-1-0.
Couldn’t be matched to a paper in Crossref, so it can’t be checked.
- Could not verifycites [14] Li, Y. et al. Single-cell transcriptome analysis reveals dynamic cell populations and differential gene expression patterns in control and aneurysmal human aortic tissue. Circulation 142, 1374–1388 (2020). sourceExperimental — accuracy not yet measured
Citation 14 is attached to individual patient variability. The Discussion describes samples from different patients and says two control sets had similar gene-expression patterns, but it does not assess Geneformer embeddings or their robustness to patient variability. Searches for gene embeddings and patient variability found no passage addressing that claim; the other two conditions are attached to different citations. (LLM, conf=0.77)
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
It may state two separate claims as one.
- Not matched
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“Depth was chosen based on the maximum depth for which there was sufficient data to pretrain as it has been established that this approach yields the greatest predictive potential in other informational fields including NLU, computer vision, and mathematical problem-solving44.”
§ Transfer learning enables predictions in network biology > Methods > Geneformer architecture and pretraining
What the sources say
- Partly supportedcites [44] Henighan, T. et al. Scaling laws for autoregressive generative modeling. Preprint at https://doi.org/10.48550/arXiv.2010.14701 (2020). sourceExperimental — accuracy not yet measured
Section 2.2 (p. 9) reports performance-maximizing depth-to-width ratios for image and math models and says these domains favor deeper models than language models. It does not establish that choosing the maximum depth permitted by available pretraining data yields the greatest predictive potential across the named fields. (LLM, conf=0.84)
boosteddirect- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: fails — The paper’s conclusions may not depend on it.
- Based: passes — The paper gives a reason why it holds.
Probably not a substantial claim. The paper’s conclusions may not depend on it. It may state two separate claims as one.
- Partly supported
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“They were predicted to be dosage-sensitive at an even lower rate in random cells from any tissue, highlighting Geneformer’s context awareness.”
§ Transfer learning enables predictions in network biology > Gene dosage sensitivity predictions
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
The quote only partly states this claim.
The paper states this without a source or data behind it here, so there is nothing to check it against. The quote shows it is in the paper.
In the paper
“Applied to disease modeling of cardiomyopathy using a limited number of patient samples, Geneformer predicted candidate therapeutic targets whose experimental targeting in an iPSC disease model led to significant functional improvement.”
§ Transfer learning enables predictions in network biology > Discussion
plaindirect- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
It may state two separate claims as one.
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“Furthermore, in silico deletion of both GATA4 and TBX5 in combination had an even greater impact on their known co-bound targets than the sum of their individual in silico deletion, suggesting Geneformer recognized their cooperative action at these co-bound targets.”
§ Transfer learning enables predictions in network biology > In silico gene network analysis
hedgeddirect- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
It may state two separate claims as one.
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“The rank value encodings for each single cell transcriptome were then tokenized based on a total vocabulary of 25,424 protein-coding or miRNA genes detected in a median of 173,152 cells within Genecorpus-30M.”
§ Transfer learning enables predictions in network biology > Methods > Assembly and rank value encoding of transcriptomes in Genecorpus-30M
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: fails — The paper’s conclusions may not depend on it.
- Based: passes — The paper gives a reason why it holds.
Probably not a substantial claim. The paper’s conclusions may not depend on it.
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“15% masking had marginally lower evaluation loss compared to 5% or 30% masking.”
§ Transfer learning enables predictions in network biology > Extended Data
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: fails — The paper’s conclusions may not depend on it.
- Based: passes — The paper gives a reason why it holds.
Probably not a substantial claim. The paper’s conclusions may not depend on it.
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“Evaluation loss was essentially equivalent in the 3 trials, indicating robustness to the set of genes randomly masked for each cell during the pretraining.”
§ Transfer learning enables predictions in network biology > Extended Data
hedged- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
It may state two separate claims as one.
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“Transcription factors are normalized by a statistically significantly lower factor (resulting in higher prioritization in the rank value encoding) compared to all genes.”
§ Transfer learning enables predictions in network biology > Extended Data
plaindirectn = 1,384- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
It may state two separate claims as one.
- The authors state this as an assumption: A representative range of transcript values would be observed within the cells in which each gene was detected.
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“Overall, 561 datasets were included and stored as uniform files in the .loom HDF5 format”
§ Transfer learning enables predictions in network biology > Methods > Assembly and rank value encoding of transcriptomes in Genecorpus-30M
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: fails — The paper’s conclusions may not depend on it.
- Based: passes — The paper gives a reason why it holds.
Probably not a substantial claim. The paper’s conclusions may not depend on it.
The paper states this without a source or data behind it here, so there is nothing to check it against. The quote shows it is in the paper.
In the paper
“The trainer from the Huggingface Transformers library159 was used for pretraining with the substitution of a custom tokenizer to implement dynamic, length-grouped padding, which minimized computation on padding and achieved a 29.4x speedup in pretraining.”
§ Transfer learning enables predictions in network biology > Methods > Geneformer architecture and pretraining
plaindirect- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: fails — The paper’s conclusions may not depend on it.
- Based: passes — The paper gives a reason why it holds.
Probably not a substantial claim. The paper’s conclusions may not depend on it.
The paper states this without a source or data behind it here, so there is nothing to check it against. The quote shows it is in the paper.
In the paper
“input size of 2048 (fully represents 93% of rank value encodings in Genecorpus-30M)”
§ Transfer learning enables predictions in network biology > Methods > Geneformer architecture and pretraining
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: fails — The paper’s conclusions may not depend on it.
- Based: fails — The paper may give no reason why it holds.
Probably not a substantial claim. The paper’s conclusions may not depend on it. The paper may give no reason why it holds.
What the paper did — it cites no source for this, so there is none to check. The quote shows it is in the paper.
In the paper
“We excluded cells with high mutational burdens (e.g. malignant cells and immortalized cell lines) that could lead to substantial network rewiring without companion genome sequencing to facilitate interpretation, and we established metrics for scalable filtering to exclude possible doublets and/or damaged cells.”
§ Transfer learning enables predictions in network biology > Geneformer architecture and pretraining
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: fails — The paper’s conclusions may not depend on it.
- Based: passes — The paper gives a reason why it holds.
Probably not a substantial claim. The paper’s conclusions may not depend on it. It may state two separate claims as one.
A conclusion the paper draws from its own results — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“A major strength of this approach is that it is entirely self-supervised and can be accomplished on completely unlabeled data, which allows the inclusion of large amounts of training data without being restricted to samples with accompanying labels.”
§ Transfer learning enables predictions in network biology > Geneformer architecture and pretraining
plaindirect- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
It may state two separate claims as one.
A conclusion the paper draws from its own results — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“Although the rank-based representation has limitations including not fully taking advantage of the precise gene expression measurements provided in transcript counts, the rank value encoding provides a nonparametric representation of each cell’s transcriptome and takes advantage of the many observations of each gene’s expression across Genecorpus-30M to prioritize genes that distinguish cell state”
§ Transfer learning enables predictions in network biology > Geneformer architecture and pretraining
plaindirect- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
It may state two separate claims as one.
A conclusion the paper draws from its own results — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“Gene regulatory network architectures are highly context-dependent; and attention-based models, known as transformers, may be exceptionally suited to context-specific modeling of network dynamics.”
§ Transfer learning enables predictions in network biology > Abstract
hedgedconditional- Asserted: passes — The authors put it forward as true.
- Checkable: fails — It may be too vague to check against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: fails — The paper may give no reason why it holds.
Probably not a substantial claim. It may be too vague to check against evidence. The paper may give no reason why it holds.
- Rests on an assumption: Gene regulatory network architectures are highly context-dependent.
A conclusion the paper draws from its own results — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“Geneformer’s ability to predict dosage-sensitive disease genes through the context-aware in silico deletion approach represents a valuable asset for interpretation of genetic variants, including prioritization of GWAS hits driving complex traits, and the specific tissues they are expected to affect.”
§ Transfer learning enables predictions in network biology > Discussion
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
It may state two separate claims as one.
In the paper
“Bivalent chromatin structure is known to mark key developmental genes in embryonic stem cells (ESCs), maintaining their promoters poised for activation28.”
§ Transfer learning enables predictions in network biology > Chromatin dynamics predictions
What the sources say
- Not checked — no full text of the sourcecites [28] Bernstein, B. E. et al. A bivalent chromatin structure marks key developmental genes in embryonic stem cells. Cell 125, 315–356 (2006). source
boosteddirect- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: fails — The paper’s conclusions may not depend on it.
- Based: passes — The paper gives a reason why it holds.
Probably not a substantial claim. The paper’s conclusions may not depend on it.
A conclusion the paper draws from its own results — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“Through the observation of a vast number of cell states during the pretraining process, Geneformer gained a fundamental understanding of network dynamics, encoding network hierarchy in the model’s attention weights in a completely self-supervised manner.”
§ Transfer learning enables predictions in network biology > Discussion
plaindirect- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
It may state two separate claims as one.
What the paper did — it cites no source for this, so there is none to check. The quote shows it is in the paper.
In the paper
“Cells with less than seven detected Ensembl-annotated protein-coding or miRNA genes were excluded as the 15% masking used for the pretraining learning objective would not reliably mask a gene in cells with fewer detected genes.”
§ Transfer learning enables predictions in network biology > Methods > Assembly and rank value encoding of transcriptomes in Genecorpus-30M
hedgeddirect- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: fails — The paper’s conclusions may not depend on it.
- Based: passes — The paper gives a reason why it holds.
Probably not a substantial claim. The paper’s conclusions may not depend on it.
A conclusion the paper draws from its own results — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“Of note, because the fine-tuning applications are trained on classification objectives that are completely separate from the masked learning objective, whether or not task-specific data was included in the pretraining corpus is not relevant to the classification predictions, as demonstrated in Extended Data Fig. 1f.”
- Also stated in: § Extended Data · plain
“Because the fine-tuning applications are trained on classification objectives that are completely separate from the masked learning objective, whether or not task-specific data was included in the pretraining corpus is not relevant to the downstream classification predictions.”
§ Transfer learning enables predictions in network biology > Methods > Geneformer fine-tuning
boosted- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
It may state two separate claims as one.
- Also stated in: § Extended Data · plain
A conclusion the paper draws from its own results — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“Overall, Geneformer represents a pretrained deep learning model whose fundamental understanding of network dynamics can now be democratized to a broad range of downstream applications to accelerate discovery of key network regulators and candidate therapeutic targets in settings with limited data.”
§ Transfer learning enables predictions in network biology > Discussion
plaindirect- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
It may state two separate claims as one.
What the paper did — it cites no source for this, so there is none to check. The quote shows it is in the paper.
In the paper
“We excluded cells with high mutational burdens (e.g. malignant cells and immortalized cell lines) that could lead to substantial network rewiring without companion genome sequencing to facilitate interpretation.”
§ Transfer learning enables predictions in network biology > Methods > Assembly and rank value encoding of transcriptomes in Genecorpus-30M
hedgeddirect- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
Probably not a substantial claim.
A conclusion the paper draws from its own results — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“It should be noted that hyperparameter tuning for deep learning applications generally significantly enhances learning and so it is likely that the maximum predictive potential of Geneformer in these downstream applications is significantly underestimated.”
§ Transfer learning enables predictions in network biology > Methods > Geneformer fine-tuning
hedgeddirect- Asserted: passes — The authors put it forward as true.
- Checkable: fails — It may be too vague to check against evidence.
- Load-bearing: fails — The paper’s conclusions may not depend on it.
- Based: fails — The paper may give no reason why it holds.
Probably not a substantial claim. It may be too vague to check against evidence. The paper’s conclusions may not depend on it. The paper may give no reason why it holds. It may state two separate claims as one.
A conclusion the paper draws from its own results — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“Generally, in our experience, applications that are more relevant to the pretraining objective benefit from more layers being frozen to prevent overfitting to the limited task-specific data, whereas applications that are more distant from the pretraining objective benefit from fine-tuning of more layers to optimize performance on the new task.”
§ Transfer learning enables predictions in network biology > Methods > Geneformer fine-tuning
hedgeddirect- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: fails — The paper’s conclusions may not depend on it.
- Based: fails — The paper may give no reason why it holds.
Probably not a substantial claim. The paper’s conclusions may not depend on it. The paper may give no reason why it holds. It may state two separate claims as one.
A conclusion the paper draws from its own results — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“A major strength of this approach is that it is entirely self-supervised and can be accomplished on completely unlabeled data, which allows the inclusion of large amounts of training data without being restricted to samples with accompanying labels.”
§ Transfer learning enables predictions in network biology > Methods > Geneformer architecture and pretraining
plaindirect- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
It may state two separate claims as one.
What the paper did — it cites no source for this, so there is none to check. The quote shows it is in the paper.
In the paper
“We only included droplet-based sequencing platforms to assure expression value unit comparability.”
§ Transfer learning enables predictions in network biology > Methods > Assembly and rank value encoding of transcriptomes in Genecorpus-30M
plaindirect- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
Probably not a substantial claim.
A conclusion the paper draws from its own results — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“We developed a novel rank value encoding method that provides a nonparametric representation of each single cell’s transcriptome, ranking genes by their expression within that cell normalized by their expression across the entire Genecorpus-30M”
§ Transfer learning enables predictions in network biology > Methods > Assembly and rank value encoding of transcriptomes in Genecorpus-30M
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
Probably not a substantial claim.
- The authors state this as an assumption: A representative range of transcript values would be observed within the cells in which each gene was detected.
What the paper did — it cites no source for this, so there is none to check. The quote shows it is in the paper.
In the paper
“Fine-tuning results for gene classification applications were reported as AUCs +/− standard deviation and F1 score calculated based on a 5-fold cross-validation strategy where training was performed on 80% of the gene training labels and performance was tested on the 20% held-out gene training labels, repeating for 5 folds.”
§ Transfer learning enables predictions in network biology > Methods > Geneformer fine-tuning
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: fails — The paper’s conclusions may not depend on it.
- Based: passes — The paper gives a reason why it holds.
Probably not a substantial claim. The paper’s conclusions may not depend on it. It may state two separate claims as one.
What the paper did — it cites no source for this, so there is none to check. The quote shows it is in the paper.
In the paper
“All fine-tuning in Supplementary Table 2 was performed with a single training epoch to avoid overfitting.”
§ Transfer learning enables predictions in network biology > Methods > Geneformer fine-tuning
plaindirect- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: fails — The paper’s conclusions may not depend on it.
- Based: passes — The paper gives a reason why it holds.
Probably not a substantial claim. The paper’s conclusions may not depend on it.
The paper relies on this without support.
Mapping gene networks requires large amounts of transcriptomic data to learn the connections between genes, which impedes discoveries in settings with limited data, including rare diseases and diseases affecting clinically inaccessible tissues.
“Mapping gene networks requires large amounts of transcriptomic data to learn the connections between genes, which impedes discoveries in settings with limited data, including rare diseases and diseases affecting clinically inaccessible tissues.”
The paper relies on this without support.
Network dynamics may vary across cell types, developmental timepoints, or disease states.
“Importantly, network dynamics may vary across cell types, developmental timepoints, or disease states.”
The paper relies on this without support.
Gene regulatory network architectures are highly context-dependent.
“Gene regulatory network architectures are highly context-dependent; and attention-based models, known as transformers, may be exceptionally suited to context-specific modeling of network dynamics.”
The paper relies on this without support.
Mapping the network hierarchy requires large amounts of transcriptional perturbation data from patient-specific cells with isogenic controls to learn the connections between genes.
“Mapping the network hierarchy required large amounts of transcriptional perturbation data from patient-specific cells with isogenic controls to learn the connections between genes.”
The paper relies on this without support.
Single cell technologies provide more precise data for inference of network interactions.
“Furthermore, single cell technologies have facilitated the observation of transcriptomic states without averaging genes’ expression across multiple cells, potentially providing more precise data for inference of network interactions, especially in diseases driven by dysregulation of multiple cell types.”
The authors state this as an assumption.
A representative range of transcript values would be observed within the cells in which each gene was detected.
“assuming that a representative range of transcript values would be observed within the cells in which each gene was detected.”
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?
The cited source doesn’t support the claim as written.
In the citing paper
“GeneFormer [189] focuses on the inference of gene function and regulatory networks, which has certain advantages in exploring the role of specific gene sets and gene clusters in cells and paves the foundation for omics-guided personalized therapy, but its generalization capability needs to be improved.”
From Challenges in AI-driven Biomedical Multimodal Data Fusion and Analysis (2025) · cites this paper as [189]
Checked by a reasoning model — may be wrong.
Experimental — accuracy not yet measuredThe source backs part of the claim, not all of it — or backs it for a different population, condition or comparison.
In the citing paper
“These models are pre-trained on vast amounts (> 10 M examples) of unlabelled scRNA-seq data, with the aim that such large datasets allow the models to capture general principles of gene regulation and signalling.”
From Benchmarking foundation cell models for post-perturbation RNA-seq prediction (2025) · cites this paper as [4]
Experimental — accuracy not yet measuredThe source backs part of the claim, not all of it — or backs it for a different population, condition or comparison.
In the citing paper
“The model outputs gene embeddings, which are subsequently decoded into gene predictions.”
From Zero-shot evaluation reveals limitations of single-cell foundation models (2025) · cites this paper as [6]
Experimental — accuracy not yet measuredThe source backs part of the claim, not all of it — or backs it for a different population, condition or comparison.
In the citing paper
“We chose these models because they offer pretrained weights and have been trained using unsupervised objectives on extensive datasets (ca. 30M single-cell transcriptomes).”
From Zero-shot evaluation reveals limitations of single-cell foundation models (2025) · cites this paper as [6]
Experimental — accuracy not yet measuredThe source backs part of the claim, not all of it — or backs it for a different population, condition or comparison.
In the citing paper
“Geneformer can output a sequence of up to 2048 genes and, when input is passed in batches, the model outputs the sequence of the length equal to that of the longest input.”
From Zero-shot evaluation reveals limitations of single-cell foundation models (2025) · cites this paper as [6]
Experimental — accuracy not yet measured
The source backs part of the claim, not all of it — or backs it for a different population, condition or comparison.
In the citing paper
“By leveraging vast datasets and pretrained transformer architectures (18), these models learn robust, generalizable relationships between genes and have demonstrated potential across diverse downstream tasks, including transcriptome prediction following perturbations after models are fine-tuned on specific datasets (19).”
From A systematic comparison of single-cell perturbation response prediction models (2026) · cites this paper as [19]
Experimental — accuracy not yet measuredThe source backs part of the claim, not all of it — or backs it for a different population, condition or comparison.
In the citing paper
“In parallel, foundation models for single-cell omics have begun to produce universal embeddings of cells and genes that are robust to batch effects and technical variability and transferable across tissues, conditions, and platforms [10, 11].”
From The next paradigm in bioinformatics: a review of multi-agent systems and foundational models for end-to-end scientific discovery (2026) · cites this paper as [10]
Experimental — accuracy not yet measuredThe source backs part of the claim, not all of it — or backs it for a different population, condition or comparison.
In the citing paper
“Both scGPT and Geneformer fine-tune their cell embeddings for cell type classification.”
From Assessing the limits of zero-shot foundation models in single-cell biology (2023) · cites this paper as [5]
Experimental — accuracy not yet measuredThe source backs part of the claim, not all of it — or backs it for a different population, condition or comparison.
In the citing paper
“We chose these models because they offer pretrained weights (whereas several other possible models did not have public weights at time of evaluation) and have been trained using unsupervised objectives on extensive datasets (ca. 30M single-cell transcriptomes).”
From Assessing the limits of zero-shot foundation models in single-cell biology (2023) · cites this paper as [5]
Experimental — accuracy not yet measuredThe source backs part of the claim, not all of it — or backs it for a different population, condition or comparison.
In the citing paper
“To evaluate their models, most previous works-including scGPT and Geneformer-rely on fine-tuning to specialize task-specific models.”
From Assessing the limits of zero-shot foundation models in single-cell biology (2023) · cites this paper as [5]
Experimental — accuracy not yet measuredThe cited source says what the claim attributes to it.
In the citing paper
“Genefomer was pretrained on 27.4M human single-cell transcriptomes (excluding malignant and immortalized cells).”
From Zero-shot evaluation reveals limitations of single-cell foundation models (2025) · cites this paper as [6]
The cited source says what the claim attributes to it.
In the citing paper
“Another limitation is that all datasets are from cancer cell lines, which, for example, Theodoris et al.5 excluded from their training data because of concerns about their high mutational burden.”
From Deep-learning-based gene perturbation effect prediction does not yet outperform simple linear baselines (2025) · cites this paper as Theodoris et al.5
The cited source says what the claim attributes to it.
In the citing paper
“The input to the Geneformer model is a ranked list where a gene’s position represents its expression relative to the remaining genes in the cell.”
From Zero-shot evaluation reveals limitations of single-cell foundation models (2025) · cites this paper as [6]
The cited source says what the claim attributes to it.
In the citing paper
“Geneformer is trained using a modification of the masked language modeling (MLM) task, where the model is trained to recover randomly selected genes that are masked or corrupted.”
From Zero-shot evaluation reveals limitations of single-cell foundation models (2025) · cites this paper as [6]
The cited source says what the claim attributes to it.
In the citing paper
“A cell embedding is calculated by averaging over all gene embeddings extracted for that cell.”
From Zero-shot evaluation reveals limitations of single-cell foundation models (2025) · cites this paper as [6]
The cited source says what the claim attributes to it.
In the citing paper
“The model leverages a BERT-inspired architecture with 6 Transformer layers, each with 4 attention heads.”
From Zero-shot evaluation reveals limitations of single-cell foundation models (2025) · cites this paper as [6]
The cited source says what the claim attributes to it.
In the citing paper
“Single-cell foundation models and perturbation predictors support cellular-response forecasting [48-51].”
From RSI-Anything: Unlocking Recursive Self-Improvement with World Models (2026) · cites this paper as [49]
In the citing paper
“This module fully retained information from the raw expression values, a notable improvement over the discretized values used in previous models 9,15 .”
From Large-scale foundation model on single-cell transcriptomics (2024) · cites this paper as [15]
Citations
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
Vaswani, A., Shazeer, N., Parmar, N., Uszkoreit, J., Jones, L., Gomez, A. N., Kaiser, L., & Polosukhin, I. (2017) · arXiv · doi:10.48550/arxiv.1706.03762 · Cited by 120,404
Jakubův, J., Chvalovský, K., Goertzel, Z., et al. (2023) · DROPS (Schloss Dagstuhl – Leibniz Center for Informatics) · doi:10.4230/lipics.itp.2023.19 · Cited by 76,648
BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding
Devlin, J., Chang, M.-W., Lee, K., & Toutanova, K. (2019) · Proceedings of the 2019 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, Volume 1 (Long and Short Papers) · doi:10.18653/v1/n19-1423 · Cited by 54,325
Decoupled Weight Decay Regularization
Loshchilov, I., & Hutter, F. (2017) · arXiv · doi:10.48550/arxiv.1711.05101 · Cited by 12,522
Fast, sensitive and accurate integration of single-cell data with Harmony
Korsunsky, I., Millard, N., Fan, J., Slowikowski, K., Zhang, F., Wei, K., Baglaenko, Y., Brenner, M., Loh, P.-R., & Raychaudhuri, S. (2019) · Nature Methods · doi:10.1038/s41592-019-0619-0 · Cited by 10,488
Analysis of protein-coding genetic variation in 60,706 humans
Exome Aggregation Consortium, Lek, M., Karczewski, K. J., Minikel, E. V., Samocha, K. E., Banks, E., Fennell, T., O’Donnell-Luria, A. H., Ware, J. S., Hill, A. J., et al. (2016) · Nature · doi:10.1038/nature19057 · Cited by 9,587
SCANPY: large-scale single-cell gene expression data analysis
Wolf, F. A., Angerer, P., & Theis, F. J. (2018) · Genome Biology · doi:10.1186/s13059-017-1382-0 · Cited by 8,881
A bivalent chromatin structure marks key developmental genes in embryonic stem cells
Bernstein, B. E., Mikkelsen, T. S., Xie, X., Kamal, M., Huebert, D. J., Cuff, J., Fry, B., Meissner, A., Wernig, M., Plath, K., et al. (2006) · Cell · doi:10.1016/j.cell.2006.02.041 · Cited by 5,361
HuggingFace's Transformers: State-of-the-art Natural Language Processing
Wolf, T., Debut, L., Sanh, V., Chaumond, J., Delangue, C., Moi, A., Cistac, P., Rault, T., Louf, R., Funtowicz, M., et al. (2019) · arXiv · doi:10.48550/arxiv.1910.03771 · Cited by 3,347
Cited by
Sorted by citations
Items 501–600 of 1,262
Wijewardena, H., Bhatia, S., Bhattacharya, N., Sengupta, D., Wu, S., & Schmitz, U. (2025) · Computational and Structural Biotechnology Journal · doi:10.1016/j.csbj.2025.11.008 · Cited by 4
Decoding cell fate: integrated experimental and computational analysis at the single-cell level
Zhou, Y., Hou, S., Miao, X., Zhang, G., Li, Z., Zhang, D., Lin, Y., & Lin, Y. (2025) · Bioinformatics · doi:10.1093/bioinformatics/btaf603 · Cited by 4
Koo, B., Sung, I., Lee, S., & Kim, S. (2025) · Briefings in Bioinformatics · doi:10.1093/bib/bbaf628 · Cited by 5
Ge, H., Mo, H., Wei, Y., Wang, J., Qi, Y., Li, L., & Ma, F. (2025) · Nature Communications · doi:10.1038/s41467-025-66384-6 · Cited by 4
Evolution of Molecular Generation Models in Drug Discovery
Xu, H.-X., Peng, J., Yu, J., Hong, X., Xiao, Y., Ye, B.-G., & Li, G.-B. (2026) · Journal of Medicinal Chemistry · doi:10.1021/acs.jmedchem.5c03310 · Cited by 4
Ngassam, B. T., Niu, H., Pang, S., Shydlouskaya, V., & Andrews, T. S. (2026) · Frontiers in Bioinformatics · doi:10.3389/fbinf.2025.1715821 · Cited by 4
Ran, R., Uslu, M., Siddiqui, M. F., Brubaker, D. K., & Trapečar, M. (2026) · Cell Reports · doi:10.1016/j.celrep.2025.116916 · Cited by 4
Improving atlas-scale single-cell annotation models with hierarchical cross-entropy loss
di Montesano, S. C., D’Ascenzo, D., Raghavan, S., Amini, A. P., Winter, P. S., & Crawford, L. A. (2026) · Nature Computational Science · doi:10.1038/s43588-025-00945-z · Cited by 4
Xia, P., Shuang, S., Fu, D., Liu, L., Yang, D., Guo, Y., Tian, Y., Ji, P., Yuan, X., Tian, Y., et al. (2026) · npj Precision Oncology · doi:10.1038/s41698-026-01307-2 · Cited by 4
Learning genetic perturbation effects with variational causal inference
Liu, E., Zhang, J., & Uhler, C. (2026) · PLoS Computational Biology · doi:10.1371/journal.pcbi.1013194 · Cited by 4
Towards predictive virtual embryos with genomics and AI
Cao, N., Lu, Y., & Qiu, X. (2026) · Nature Methods · doi:10.1038/s41592-026-03055-4 · Cited by 5
Shen, H., Zhang, L., Ma, X., Yin, Y., Wang, J., & Tan, B. (2026) · Science China Life Sciences · doi:10.1007/s11427-025-3163-6 · Cited by 5
Target discovery and drug design in the era of artificial intelligence
Fleming, P., & Ivanov, A. A. (2026) · Medicinal Chemistry Research · doi:10.1007/s00044-026-03562-1 · Cited by 4
Discovery of candidate therapeutic targets with Geneformer
Zhang, Y., Venkatesh, M. S., & Theodoris, C. V. (2026) · Nature Protocols · doi:10.1038/s41596-026-01364-8 · Cited by 5
Self-citationCell-o1 : training LLMs to solve single-cell reasoning puzzles with reinforcement learning
Fang, Y., Jin, Q., Xiong, G., Jin, B., Zhong, X., Ouyang, S., Yang, Y., Zhang, A., Han, J., & Lu, Z. (2026) · Bioinformatics · doi:10.1093/bioinformatics/btag208 · Cited by 12
Kendiukhov, I. (2026) · Biosystems · doi:10.1016/j.biosystems.2026.105864 · Cited by 4
Liu, G., Wang, J., & Zhu, J. (2026) · Global Health Care · doi:10.63808/ghc.v2i3.498 · Cited by 4
Lin, H., Makhlouf, A., Echegaray, C. V., Zawada, D. M., & Simões, F. C. (2023) · Development · doi:10.1242/dev.202300 · Cited by 4
85 million cells — and counting — at your fingertips
Perkel, J. M. (2024) · Nature · doi:10.1038/d41586-024-01217-y · Cited by 4
Gene pointNet for tumor classification
Lü, H., Rezapour, M., Baha, H., Niazi, M. K. K., Narayanan, A., & Gürcan, M. N. (2024) · Neural Computing and Applications · doi:10.1007/s00521-024-10307-x · Cited by 4
Spatially aligned graph transfer learning for characterizing spatial regulatory heterogeneity
Huang, W., Hu, Y., Wang, L., Wu, G., Zhang, C., & Shi, Q. (2024) · Briefings in Bioinformatics · doi:10.1093/bib/bbaf021 · Cited by 4
Unified Integration of Spatial Transcriptomics Across Platforms
Haber, E., Deshpande, A., Ma, J., & Krieger, S. (2025) · Lecture notes in computer science · doi:10.1007/978-3-031-90252-9_42 · Cited by 4
Zhao, J., Zhou, Y., Bao, H., Zhao, X., Wang, X., Zhao, C., Qin, W., Lu, X., & Xu, G. (2025) · Briefings in Bioinformatics · doi:10.1093/bib/bbaf532 · Cited by 4
UniCure: A multi-modal model for predicting personalized cancer therapy response
Chen, Z., Tian, S., Pei, J., Gu, R., Li, Y., Ding, S., Xu, Y., Zheng, X., Liu, M., Du, X., et al. (2026) · Cancer Cell · doi:10.1016/j.ccell.2026.07.010 · Cited by 4
Giordano, M., Falbo, E., Maddalena, L., Piccirillo, M., & Granata, I. (2023) · Biomolecules · doi:10.3390/biom14010018 · Cited by 4
A deep learning framework for automated and generalized synaptic event analysisPreprint
O’Neill, P. S., Baccino-Calace, M., Rupprecht, P., Friedrich, R. W., Müller, M., & Delvendahl, I. (2024) · eLife · doi:10.7554/elife.98485.1 · Cited by 4
CSLAN: Cross-Species Latent Alignment Network for Trauma-Related Cell-Type ClassificationPreprint
Wu, R., Shi, A., Liu, Y., Duprey, A., Chen, T., Li, R., & Cao, S. (2025) · bioRxiv (Cold Spring Harbor Laboratory) · doi:10.1101/2025.06.09.655966 · Cited by 4
Bahrami, M., Tejada-Lapuerta, A., Becker, S., G., F. S. H., & Theis, F. J. (2025) · bioRxiv (Cold Spring Harbor Laboratory) · doi:10.1101/2025.10.14.682419 · Cited by 4
TwinCell: Large Causal Cell Model for Reliable and Interpretable Therapeutic Target PrioritisationPreprint
Morlot, J.-B., Dias, T. L., Légaré, S., Romualdi, A., Hatem, E., & Abraham, Y. (2026) · bioRxiv (Cold Spring Harbor Laboratory) · doi:10.64898/2026.01.29.702072 · Cited by 4
annATAC: automatic cell type annotation for scATAC-seq data based on language model
Cui, L., Wang, F., Li, H., Liu, Q., Zhou, M., & Wang, G. (2025) · BMC Biology · doi:10.1186/s12915-025-02244-5 · Cited by 3
GenoHoption: Bridging Gene Network Graphs and Single-Cell Foundation Models
Cheng, J., Li, J., Yang, K., Shen, H., & Yuan, Y. (2024) · IEEE International Conference on Bioinformatics and Biomedicine (BIBM) · doi:10.1109/bibm62325.2024.10822153 · Cited by 3
Wang, J., Huang, Y., & Winther, O. (2026) · Nature Computational Science · doi:10.1038/s43588-026-01016-7 · Cited by 3
Zhang, Q., Wu, X., Li, X., Ma, W., Wu, T., Li, L., Hu, F., Xie, Y., & Wu, X. (2025) · Computer Methods and Programs in Biomedicine · doi:10.1016/j.cmpb.2025.108809 · Cited by 3
Multi-domain fake news detection method based on generative adversarial network and graph network
Li, X., Wei, J., Zhao, C., Fan, X., & Wang, Y. (2025) · Knowledge-Based Systems · doi:10.1016/j.knosys.2025.113665 · Cited by 3
Zhang, H., Jiang, Z., Zhang, S., Tu, L., & Carlson, D. (2025) · Pattern Recognition · doi:10.1016/j.patcog.2025.111724 · Cited by 3
FR-BINN: Biologically Informed Neural Networks for Enhanced Biomarker Discovery and Pathway Analysis
Cao, Y., Yin, C., Zhou, X., & Zhao, Y. (2025) · International Journal of Molecular Sciences · doi:10.3390/ijms26146670 · Cited by 3
GeneRAIN: multifaceted representation of genes via deep learning of gene expression networks
Su, Z., Fang, M., Smolnikov, A., Dinger, M. E., Oates, E. C., & Vafaee, F. (2025) · Genome biology · doi:10.1186/s13059-025-03749-6 · Cited by 3
Zhang, Y., Wei, R., Wei, S., Zheng, X., & Long, H. (2025) · Energy Conversion and Management · doi:10.1016/j.enconman.2025.120643 · Cited by 3
A novel paradigm for single-cell annotation in stem cell research
Atkeson, T., Wilson, S. B., Ruiz-Morales, E. R., Little, M. H., Vento‐Tormo, R., Neavin, D. R., & Powell, J. E. (2025) · Stem Cell Reports · doi:10.1016/j.stemcr.2025.102707 · Cited by 3
Mou, J., Yan, Y., Jiang, B., Yang, F., Pan, Z., Huang, X., Bai, M., Han, Z., & Li, Y. (2025) · Molecular Diversity · doi:10.1007/s11030-025-11394-1 · Cited by 3
Enhancing key node identification in hypernetworks: A novel node–hyperedge interaction model
Yan, F., Liu, S., Tang, Y., Witold Pedrycz, C., & Hirota, K. (2025) · Chaos Solitons & Fractals · doi:10.1016/j.chaos.2025.117390 · Cited by 3
An expanded role for single-cell chemical genomics profiling in drug discovery
Wyatt, A., Hoffer-Hawlik, K., Giglio, R. M., Azizi, E., & McFaline‐Figueroa, J. L. (2026) · Biochemical Journal · doi:10.1042/bcj20253273 · Cited by 3
Yu, Z., Yang, D., Chen, Q., Zhang, Y., Li, Z., Wang, Y., Wang, C., & Zhang, Y. (2026) · Cell Genomics · doi:10.1016/j.xgen.2025.101130 · Cited by 4
Model Synthesizer: Synthesizing Disparate Models to Enhance Single-Cell Transcriptomics Models
Chen, S., Cao, Z., Qin, X. Y., & Yang, X. W. (2025) · IEEE International Conference on Bioinformatics and Biomedicine (BIBM) · doi:10.1109/bibm66473.2025.11356145 · Cited by 3
Machine Learning‐Guided Engineering of Protein Phase Separation Properties in Immune Regulation
Zhang, C., Wang, J., Wang, Z., Zhu, L., Cai, S., Zhan, L., Liang, H., Wu, Y., Li, J., & Cui, J. (2026) · Advanced Science · doi:10.1002/advs.202520890 · Cited by 3
Liu, Z., Jiang, S., Hu, T., Xue, Z., Chen, Y., Ruan, D., Xia, H., Feng, Y., Hao, J., Li, W., et al. (2026) · Patterns · doi:10.1016/j.patter.2025.101471 · Cited by 4
Non-viral and mRNA-based strategies for next-generation CAR-T cell therapy
Wang, T., Jin, F., Xing, W., & Yu, S. (2026) · Pathology - Research and Practice · doi:10.1016/j.prp.2026.156455 · Cited by 3
From foundation models to autonomous agents in biology
Huang, S., Lang, M., Chen, Z., Yang, C., Huang, X., Mohtashaminia, Z., & Peng, Y. (2026) · Genomics Communications · doi:10.48130/gcomm-0026-0005 · Cited by 4
MHGNN: Multiplex Hypergraph Neural Networks for Predicting Herb--Symptom Interactions
Liang, X., Lin, T., Xie, B., Tang, Y., & Wang, W. (2026) · IEEE Transactions on Neural Networks and Learning Systems · doi:10.1109/tnnls.2026.3677056 · Cited by 3
Nuanpirom, J., & Charoensawan, V. (2026) · Briefings in Bioinformatics · doi:10.1093/bib/bbag152 · Cited by 3
Spatial biomarker discovery via interpretable semantic learning in histopathology
Liang, J., Jiang, X., Reitsam, N. G., Lenz, T., Zhang, L., Gustav, M., Carrero, Z. I., Muti, H. S., Neidlinger, P., Clusmann, J., et al. (2026) · Cancer Cell · doi:10.1016/j.ccell.2026.05.014 · Cited by 3
Mapping and engineering the human cell–cell interactome
Carlo, D. D., Morsut, L., McCain, M. L., Wright, H. J., Abedi, M. H., Yamada‐Hunter, S. A., Zhang, J. Z., Backus, K. M., Chung, E., Wang, Y., et al. (2026) · Nature Biotechnology · doi:10.1038/s41587-026-03177-2 · Cited by 3
Beyond static biomarkers: systems biology and AI for decoding cancer dynamics
Dutta, S., & Goli, A. (2026) · Frontiers in Systems Biology · doi:10.3389/fsysb.2026.1855016 · Cited by 3
Decoding Cardiac Development and Maturation at Single-Cell and Spatial Transcriptomic Resolution
Li, B., Liu, X., Murphy, S. A., Tampakakis, E., & Kwon, C. (2026) · Circulation Research · doi:10.1161/circresaha.125.327473 · Cited by 3
Self-citationMacaulay, O., Jararweh, A., Arredondo, D., Tafoya, L. E., Poshiya, S., Hu, Y., Oyebamiji, O., Zhang, Y., Amos, C. I., Virupakshappa, K., et al. (2026) · Nature Communications · doi:10.1038/s41467-026-78454-4 · Cited by 3
LWTD: a novel light-weight transformer-like CNN architecture for driving scene dehazing
Zhang, Z., Feng, Z., Long, A., & Wang, Z. (2024) · International Journal of Machine Learning and Cybernetics · doi:10.1007/s13042-024-02335-9 · Cited by 3
DeepMBEnzy: An AI-Driven Database of Mycotoxin Biotransformation Enzymes
Cai, P., Liu, D., Xing, H., Zhang, D., Le, Y., Wu, A., & Hu, Q. (2025) · Journal of Agricultural and Food Chemistry · doi:10.1021/acs.jafc.5c02477 · Cited by 3
A perspective on developing foundation models for analyzing spatial transcriptomic data
Liu, T., Hao, M., Liu, X., & Zhao, H. (2025) · Quantitative Biology · doi:10.1002/qub2.70010 · Cited by 3
Self-citationFoundation Model: A New Era for Plant Single-cell Genomics
Zeng, Y., & Yang, Y. (2025) · Genomics Proteomics & Bioinformatics · doi:10.1093/gpbjnl/qzaf059 · Cited by 3
Deng, D., Xu, W., Wu, B., Comes, H. P., Feng, Y. Z., Li, P., Zheng, J., Chen, G., & Heng, P. (2025) · Nature Communications · doi:10.1038/s41467-025-61684-3 · Cited by 3
The cell as a token: high-dimensional geometry in language models and cell embeddings
Gilpin, W. (2025) · Bioinformatics · doi:10.1093/bioinformatics/btaf595 · Cited by 3
Liang, H., Berger, B., & Singh, R. (2026) · Cell Systems · doi:10.1016/j.cels.2026.101675 · Cited by 3
Towards foundation models that learn across biological scales
Kalfon, J., Cantini, L., & Gabriel Peyré, A. (2026) · Machine Learning Health · doi:10.1088/3049-477x/aea764 · Cited by 4
AI learns from chromatin data to uncover gene interactions
Brożek, A., & Theodoris, C. V. (2025) · Nature · doi:10.1038/d41586-024-04107-5 · Cited by 3
Self-citationTang, Z., Wang, F., Yang, F., Song, J., Li, Y., Zhou, J., Song, Y., Chen, S., Zhu, J., You, L., et al. (2026) · Nature Communications · doi:10.1038/s41467-026-77538-5 · Cited by 4
Scaling deep identifiable models enables zero-shot characterization of single-cell biological statesPreprint
Dong, M., Agrawal, K., Fan, R., Sefik, E., Flavell, R. A., & Kluger, Y. (2023) · bioRxiv (Cold Spring Harbor Laboratory) · doi:10.1101/2023.11.11.566161 · Cited by 3
Liang, Y., Li, D., Xu, A. G., Shao, Y., & Tang, K. (2024) · bioRxiv (Cold Spring Harbor Laboratory) · doi:10.1101/2024.06.27.601098 · Cited by 3
SC-MAMBA2: Leveraging State-Space Models for Efficient Single-Cell Ultra-Long Transcriptome ModelingPreprint
Zhao, Y., Zhao, B., Zhang, F., He, C., Wu, W., & Lai, L. (2024) · bioRxiv (Cold Spring Harbor Laboratory) · doi:10.1101/2024.09.30.615775 · Cited by 3
scPlantFormer: A Lightweight Foundation Model for Plant Single-Cell Omics AnalysisPreprint
Zhang, X., Xu, J., Chen, D., & Chen, L. (2024) · Research Square · doi:10.21203/rs.3.rs-5219487/v1 · Cited by 3
Towards a Generative Paradigm for Large-scale Microbiome Analysis by Generative Language ModelPreprint
Zhang, H., Kang, Z., Zhang, Y., Yang, R., & Ning, K. (2025) · bioRxiv (Cold Spring Harbor Laboratory) · doi:10.1101/2025.01.15.633278 · Cited by 3
Hu, H., Zhang, S., Choi, Y., Malladi, V. S., & Quon, G. (2025) · bioRxiv (Cold Spring Harbor Laboratory) · doi:10.1101/2025.01.28.635153 · Cited by 3
Evaluating Foundation Models for In-Silico PerturbationPreprint
Liu, X., Boylan, J., Bouiller, T., Solovyeva, E. M., Ataman, B., Hoersch, S., Jenkins, J. L., & Devarakonda, M. V. (2025) · bioRxiv (Cold Spring Harbor Laboratory) · doi:10.1101/2025.05.11.653338 · Cited by 4
Self-citationHe, C., Zhang, J., Dahleh, M. A., & Uhler, C. (2025) · bioRxiv (Cold Spring Harbor Laboratory) · doi:10.1101/2025.06.27.661992 · Cited by 4
Closing the loop: Teaching single-cell foundation models to learn from perturbationsPreprint
Pershad, Y., Nandi, T., Amburg, J. C. V., Parker, A., Ostrowski, L., Giannini, H. K., Ong, D. E., Heimlich, J. B., Obeng, E. A., Ericson, K. R., et al. (2025) · bioRxiv (Cold Spring Harbor Laboratory) · doi:10.1101/2025.07.08.663754 · Cited by 3
Decoding the gene regulatory landscape through multimodal learning of protein-DNA interactionsPreprint
Tan, J., Fu, X., Ling, X., Mo, S., Bai, J., Rabadán, R., Fenyö, D., Boeke, J. D., Tsirigos, A., & Xia, B. (2025) · bioRxiv (Cold Spring Harbor Laboratory) · doi:10.1101/2025.08.17.670761 · Cited by 5
GeneJepa: A Predictive World Model of the TranscriptomePreprint
Litman, E., Myers, T. D., Agarwal, V., Mittal, E., Li, O., Gopinath, A., & Kassis, T. (2025) · bioRxiv (Cold Spring Harbor Laboratory) · doi:10.1101/2025.10.14.682378 · Cited by 4
Empirical Evaluation of Single-Cell Foundation Models for Predicting Cancer OutcomesPreprint
Roman, A., Johri, S., Conci, R., Allen, E. M. V., & Elmarakeby, H. (2025) · bioRxiv (Cold Spring Harbor Laboratory) · doi:10.1101/2025.10.31.685892 · Cited by 3
Atacformer: A transformer-based foundation model for analysis and interpretation of ATAC-seq dataPreprint
LeRoy, N. J., Zheng, G., Khoroshevskyi, O., Campbell, D. R., Zhang, A., & Sheffield, N. C. (2025) · bioRxiv (Cold Spring Harbor Laboratory) · doi:10.1101/2025.11.03.685753 · Cited by 3
Unified integration of spatial transcriptomics across platforms with LLOKIPreprint
Haber, E., Deshpande, A., Ma, J., & Krieger, S. (2025) · Genome Research · doi:10.1101/gr.280803.125 · Cited by 3
Language may be all omics needs: Harmonizing multimodal data for omics understanding with CellHermesPreprint
Gao, Y., Wang, W., Zhao, Y., Dong, K., Shan, C., Zheng, W., Richter, T., Li, Z., Chen, S., Theis, F. J., et al. (2025) · bioRxiv (Cold Spring Harbor Laboratory) · doi:10.1101/2025.11.07.687322 · Cited by 3
mcBERT: Patient-Level Single-cell Transcriptomics Data RepresentationPreprint
von Querfurth, B., Lohmöller, J., Pennekamp, J., Bleckwehl, T., Kramann, R., Wehrle, K., & Hayat, S. (2024) · bioRxiv (Cold Spring Harbor Laboratory) · doi:10.1101/2024.11.04.621897 · Cited by 3
In-Context Learning can distort the relationship between sequence likelihoods and biological fitnessPreprint
Kantroo, P., Wagner, G. P., & Machta, B. B. (2025) · PubMed · doi:10.48550/arxiv.2504.17068 · Cited by 3
Evaluating Single-Cell Perturbation Response Models Is Far from StraightforwardPreprint
Heidari, M., Karimpour, M., Srivatsa, S., & Montazeri, H. (2026) · bioRxiv (Cold Spring Harbor Laboratory) · doi:10.64898/2026.02.14.705879 · Cited by 5
Song, H., Xu, J., Velázquez-Arcelay, K., Demirci, A., Raizenne, B. L., Hsu, S. C., Choi, J., Pham, J. H., Chen, Y.-A., Weinstein, H. N. W., et al. (2026) · bioRxiv (Cold Spring Harbor Laboratory) · doi:10.64898/2026.03.25.711335 · Cited by 3
Aniseia, Y., Waltari, E., HUANG, H. H., Lima, L., Rahman, G., Frank, M., Zhou, A., Kim, Y. J., Paras, J., Baker, S. F., et al. (2026) · bioRxiv (Cold Spring Harbor Laboratory) · doi:10.64898/2026.06.17.732950 · Cited by 3
Tabular Foundation Models Are Competitive Cellular Perturbation Predictors Across Biological ScalesPreprint
Palla, G., Hillsley, A. V., Kim, Y. J., & Royer, L. A. (2026) · bioRxiv (Cold Spring Harbor Laboratory) · doi:10.64898/2026.06.28.735106 · Cited by 3
Branda, F., Ahmed, M. M., Ciccozzi, M., Guzzi, P. H., & Scarpa, F. (2026) · Briefings in Bioinformatics · doi:10.1093/bib/bbag245 · Cited by 2
A self-training interpretable cell type annotation framework using specific marker gene
Chen, H., Lu, Y., & Rao, Y. (2024) · Bioinformatics · doi:10.1093/bioinformatics/btae569 · Cited by 3
Luo, M., Wang, C., & Lee, S. (2024) · Optics Express · doi:10.1364/oe.540023 · Cited by 2
Unveiling Long Non-coding RNA Networks from Single-Cell Omics Data Through Artificial Intelligence
Cao, G., & Chen, D. (2024) · Methods in molecular biology · doi:10.1007/978-1-0716-4290-0_11 · Cited by 2
Fan, X., Xiang, G., Liao, W., Xiao, L., He, S., Luo, N., Lu, H., & Yan, X. (2025) · Synthetic and Systems Biotechnology · doi:10.1016/j.synbio.2025.06.005 · Cited by 2
Self-citationJia, S., Liang, S., Mo, Z., Liu, C., Wang, H., & Chen, C. (2025) · Expert Systems with Applications · doi:10.1016/j.eswa.2025.129406 · Cited by 3
Motor bearing fault diagnosis based on industrial internet of things and transfer learning
Guo, D. (2025) · Frontiers in Mechanical Engineering · doi:10.3389/fmech.2025.1647310 · Cited by 2
Wang, X., Mei, Y., Wang, X., Wu, H., Wang, R., Zhang, P., Zhang, G., Li, J., Jiang, M., Fang, X., et al. (2025) · Clinical and Translational Medicine · doi:10.1002/ctm2.70461 · Cited by 3
Artificial intelligence for drug target identification
Li, Y., Huang, J., Sheng, Y., Wei, C., Liu, M., Wang, X., Chen, Y., Li, X., & Zheng, M. (2025) · Annual reports in medicinal chemistry · doi:10.1016/bs.armc.2025.08.002 · Cited by 2
An Edge Cloud Collaborative Leakage Detection Approach for Urban Water Distribution Systems
Qiao, J., Xiao, C., & Fu, A. (2025) · IEEE Transactions on Instrumentation and Measurement · doi:10.1109/tim.2025.3613911 · Cited by 2
Zhang, X., Luo, J., Pan, W., Xue, Q., Liu, X., Fu, J., Zhang, A., & Jiang, G. (2025) · Atmospheric chemistry and physics · doi:10.5194/acp-25-13379-2025 · Cited by 2
Self-citationXu, G., Wang, Z., Zhou, J., Yu, W., & Ou-Yang, L. (2025) · IEEE Journal of Biomedical and Health Informatics · doi:10.1109/jbhi.2025.3628564 · Cited by 3
scProtoTransformer: Scalable reference mapping across molecules, cells, and donors
Tang, Z., He, H., Chen, S., Zhu, J., Lv, T., Jiale, Z., Huang, J., Li, Y., Chen, G., You, L., et al. (2026) · Science Advances · doi:10.1126/sciadv.aef0286 · Cited by 3
Artificial intelligence in radiobiology: Bridging mechanisms and data analysis
Wang, Y., Li, W., Xu, N., Xie, D., Huang, X., & Zhou, P. (2025) · Radiation Medicine and Protection · doi:10.1016/j.radmp.2025.12.005 · Cited by 2
Metadata
- Authors
- Theodoris, C. V., Xiao, L., Chopra, A., Chaffin, M. D., Al Sayed, Z. R., Hill, M. C., Mantineo, H., Brydon, E. M., Zeng, Z., Liu, X. S., & Ellinor, P. T.
- Affiliations
- Broad Institute (US)
- Harvard University (US)
- Corresponding author
- Theodoris, Christina V.
- Ellinor, Patrick T.
- ORCID
- Venue
- Nature
- Published
- 2023-05-31
- Volume
- 618
- Issue
- 7965
- Pages
- 616-624
- ISSN
- 0028-0836 (ISSN-L)
- 1476-4687
- PMC
- PMC10949956
- OpenAlex
- W4378838672
- Cited by
- 1,193
- References
- 157
- Citations by year
- 2026: 488 · 2025: 402 · 2024: 257 · 2023: 46
- FWCI
- 105.89
- Type
- journal article
- Language
- en
- Open access
- green
- www.ncbi.nlm.nih.gov
- Licence
- Locations
- Nature (publishedVersion)
- PubMed (publishedVersion)
- PubMed Central (acceptedVersion)
- eScholarship (California Digital Library) (submittedVersion)
- RePEc: Research Papers in Economics (submittedVersion)
- Funders
- American Heart Association (1R01HL092577, 18SFRN34110082, 965286, MAESTRIA 965286, 20CDA35260081)
- Bristol-Myers Squibb
- Pfizer
- Bayer
- MyoKardia
- European Commission (965286)
- National Institutes of Health (965,286, 18SFRN34110082)
- National Heart, Lung, and Blood Institute (R01HL092577, R01HL157635, R01HL139731)
- National Institute of General Medical Sciences (T32GM007748)
- Topics
- Life Sciences › Biochemistry, Genetics and Molecular Biology › Molecular Biology › Bioinformatics and Genomic Networks
- Life Sciences › Biochemistry, Genetics and Molecular Biology › Molecular Biology › Gene Regulatory Network Analysis
- Life Sciences › Biochemistry, Genetics and Molecular Biology › Molecular Biology › Single-cell and spatial transcriptomics
- Keywords
- transfer learning, Geneformer, gene regulatory networks, single-cell transcriptomics, attention-based deep learning, self-supervised pre-training, chromatin dynamics, cardiomyopathy, therapeutic target discovery, network biology
- MeSH terms
- Machine Learning, Single-Cell Gene Expression Analysis, Biology, Chromatin, Humans, Cardiomyopathies, Neural Networks, Computer, Datasets as Topic
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.
A language model drew the claims from the paper on 2026-10-20, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).
A second model labelled each claim on 2026-10-20 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).
Cited sources are checked with a language model’s help (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
