The DOI matches its Crossref record; 127 of the 127 references it cites were found in a registry or an index; 32 of the 49 sources its claims cite were checked for support.
Few shot learning for phenotype-driven diagnosis of patients with rare genetic diseases
Open accessPublished version of a preprint
Abstract
There are over 7000 rare diseases, some affecting 3500 or fewer patients in the United States. Due to clinicians' limited experience with such diseases and the heterogeneity of clinical presentations, ~70% of individuals seeking a diagnosis remain undiagnosed. Deep learning has demonstrated success in aiding the diagnosis of common diseases. However, existing approaches require labeled datasets with thousands of diagnosed patients per disease. We present SHEPHERD, a few-shot learning approach for multi-faceted rare disease diagnosis. SHEPHERD performs deep learning over a knowledge graph enriched with rare disease information and is trained on a dataset of simulated rare disease patients. We demonstrate SHEPHERD's effectiveness across diverse diagnostic tasks, performing causal gene discovery, retrieving "patients-like-me", and characterizing novel disease presentations, using real-world cohorts from the Undiagnosed Diseases Network (N = 465), MyGene2 (N = 146), and the Deciphering Developmental Disorders study (N = 1431). SHEPHERD demonstrates the potential of knowledge-grounded deep learning to accelerate rare disease diagnosis.
Overview
CoChat Summary
TL;DR
SHEPHERD is a few-shot geometric deep learning method that uses a rare disease knowledge graph and simulated patients to perform phenotype-driven causal gene discovery, “patients-like-me” retrieval, and novel disease characterization. Evaluated on real cohorts including 465 Undiagnosed Diseases Network patients, it often ranks causal genes highly and generalizes across sites, diseases, and atypical presentations.
Key findings
- On UDN EXPERT-CURATED gene lists, SHEPHERD ranked the causal gene first in 40% of patients, with average recall 0.69 at k = 3 and 0.85 at k = 5; on longer VARIANT-FILTERED lists, average recall was 0.21, 0.38, and 0.48 at k = 1, 5, and 10.
- For hard-to-diagnose patients whose phenotype terms were more than two hops from the causal gene in the knowledge graph, SHEPHERD identified the correct causal gene among its top five predictions for 77.8% of patients.
- In hard scenarios, SHEPHERD achieved win rates up to 82% and 83% for patients whose causal genes lacked known phenotype or disease associations, respectively, on EXPERT-CURATED lists, and up to 86% on VARIANT-FILTERED lists for novel disease genes.
- SHEPHERD’s patient embeddings captured disease similarity (adjusted mutual information = 0.304; p < 0.01) and enabled “patients-like-me” retrieval, ranking a MyGene2 patient with the same causal gene first for 11.5% of UDN patients and achieving recall 0.31, 0.43, 0.49, and 0.53 at k = 5, 10, 25, and 50.
Methods
- SHEPHERD uses a rare disease knowledge graph with 105,220 nodes and 1,095,469 edges, built from sources including PrimeKG, HPO, MONDO, Orphanet, DisGeNET, GO, and Reactome.
- It is pretrained via self-supervised link prediction on the graph, then fine-tuned for diagnosis tasks using a graph attention network, patient phenotype subgraphs, and disease-stratified training on 42,624 simulated patients representing 2,132 rare diseases.
- Evaluation used real cohorts: UDN (N = 465), MyGene2 (N = 146), and DDD (N = 1431), with causal gene discovery compared against domain-specific tools, language models, traditional machine learning, network science baselines, and random ranking.
Limitations
- The paper states that continually updating the knowledge graph could improve performance, that still-undiagnosed UDN patients may be more challenging than diagnosed patients tested, and that datasets likely overrepresent individuals of European descent, limiting generalizability.
- The text also makes evident that evaluation is retrospective on already-diagnosed patients and does not assess prospective clinical workflow impact; the authors note prospective evaluation as an important next step.
- The paper states that SHEPHERD relies on structured phenotype terms and a knowledge graph, and suggests extensions such as incorporating variant-level information, negative phenotypes, temporal modeling, and multimodal data.
Model-written summary. Check it against the paper before relying on it.
Integrity signals
Registry
Cites
Statistics
Figures






Verification
Claims the paper makes
20 Valid · 5 Not valid · 6 Not checked · 194 Own results and interpretations
At least one cited source doesn’t support this claim or points to a different paper, and none confirmed it — see the verdicts below.
In the paper
“a gene previously unassociated with any disease61”
§ Few shot learning for phenotype-driven diagnosis of patients with rare genetic diseases > Results > SHEPHERD provides an interpretable characterization of novel diseases
What the sources say
- Disputedcites [61] Ganapathi, M. et al. A homozygous splice variant in atp5po, disrupts mitochondrial complex v function and causes leigh syndrome in two unrelated families. J. Inherit. Metab. Dis. 45, 996–1012 (2022). sourceExperimental — accuracy not yet measured
The Introduction (page unavailable) says a recent report had already described an affected individual with compound heterozygous ATP5PO variants. This contradicts the claim that ATP5PO had previously been unassociated with any disease. (LLM, conf=0.89)
Checked by a reasoning model — may be wrong.
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.
- Disputed
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“leveraging indirect associations is essential for diagnosing patients with novel or atypical genetic conditions”
§ Few shot learning for phenotype-driven diagnosis of patients with rare genetic diseases > Discussion
What the sources say
- Partly supportedcites [16] Jagadeesh, K. A. et al. Phrank measures phenotype sets similarity to greatly improve Mendelian diagnostic disease prioritization. Genet. Med. 21, 464–470 (2019). sourceExperimental — accuracy not yet measured
The Methods section on Phrank inputs (p. 3) describes using ontology relationships to infer broader phenotype associations from specific ones, supporting the idea of leveraging indirect associations for diagnostic ranking. But the paper does not establish that doing so is essential for diagnosing novel or atypical conditions; its benchmark section (p. 4) says cases proposing novel causative genes were excluded. (LLM, conf=0.86)
- Not checked — no full text of the sourcecites [18] Köhler, S. et al. Clinical diagnostics in human genetics with semantic similarity searches in ontologies. Am. J. Hum. Genet. 85, 457–464 (2009). source
plainstates a cause- 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
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“Training on simulated data mitigates concerns regarding privacy breaches, in which specific individuals can be identified from the training data82,83.”
§ Few shot learning for phenotype-driven diagnosis of patients with rare genetic diseases > Discussion
What the sources say
- Partly supportedcites [82] Oprisanu, B., Ganev, G. & De Cristofaro, E. On utility and privacy in synthetic genomic data. In Proceedings of the 29th Network and Distributed System Security Symposium (NDSS, 2022). sourceExperimental — accuracy not yet measured
The Introduction describes synthetic data as a way to minimize the risk of inferring sensitive information about individuals. But Main Findings says synthetic datasets only sometimes offer better protection against membership inference and do not provide robust protection; the claim omits that important limitation. (LLM, conf=0.86)
- Supportedcites [83] Wang, Z., Myles, P. & Tucker, A. Generating and evaluating cross-sectional synthetic electronic healthcare data: preserving data utility and patient privacy. Comput. Intell. 37, 819–851 (2021). source
The Introduction says that patient identification is more likely with healthcare data and that a fully synthetic approach can better mitigate this risk. Section 2.2 adds that synthetic data can tackle privacy concerns, while noting that disclosure risk can remain if synthetic outliers match real patients. (LLM, conf=0.91)
plainstates a cause- 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
“The simulated patients closely resemble real-world patients found in the UDN.”
§ Few shot learning for phenotype-driven diagnosis of patients with rare genetic diseases > Simulated patients with rare Mendelian disorders
What the sources say
- Supportedcites [48] Alsentzer, E. et al. Simulation of undiagnosed patients with novel genetic conditions. Nat. Commun. 14, 6403 (2023). source
The Results section, “Simulated patients mimic real-world patients,” directly compares disease-matched simulated patients with diagnosed UDN patients. It reports similar numbers of candidate genes and positive phenotype terms, and says the groups cluster together within disease categories. The comparison applies to the 121 UDN patients whose diseases the authors could model. (LLM, conf=0.94)
plainstates an association- 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
“A notable byproduct of training the model on synthetic data is that SHEPHERD’s model can be publicly released without the risk of exposing patient information52.”
§ Few shot learning for phenotype-driven diagnosis of patients with rare genetic diseases > Results > Overview of SHEPHERD algorithm
What the sources say
- Partly supportedcites [52] Chen, R. J., Lu, M. Y., Chen, T. Y., Williamson, D. F. K. & Mahmood, F. Synthetic data in machine learning for medicine and healthcare. Nat. Biomed. Eng. 5, 493–497 (2021). sourceExperimental — accuracy not yet measured
The Fidelity tests section says synthetic datasets could be freely shared and would substantially lessen patient-privacy concerns. It does not establish that SHEPHERD’s model can be released without risk; the same section warns that publicly available model weights can leak patient information. (LLM, conf=0.84)
plainstates a causen = 40,000- 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.
- Partly supported
122 more claims · 8 with their cited sources checked
Create a free account to see all 225 claims and to verify the 4 references still open.
- It may state two separate claims as one. 63
- Probably not a substantial claim. 43
- The quote only partly states this claim. 1
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 source backs part of the claim, not all of it — or backs it for a different population, condition or comparison.
In the citing paper
“Early research has shown that LLMs can generate synthetic patient data to augment real datasets for rare conditions [44].”
From A Review of Multi-Agent AI Systems for Biological and Clinical Data Analysis (2026) · cites this paper as [44]
Experimental — accuracy not yet measured
Self-check
Percentages, sums and reported differences re-computed from the paper’s own numbers.
Each contradiction between two passages listed here was confirmed by an independent check.
Rated minor: it would not touch a result — a value no result rests on, a p-value that stays on the same side of the significance level, or a conflict that is clerical, resolved elsewhere in the paper, or outside its own findings.
In the paper
§ Few shot learning for phenotype-driven diagnosis of patients with rare genetic diseases > Results > Overview of the undiagnosed diseases network patient cohort
“UDN patients have heterogeneous disease presentations: 378 unique genes and 299 unique diseases are represented in the cohort”
§ Few shot learning for phenotype-driven diagnosis of patients with rare genetic diseases > Patients in the Undiagnosed Diseases Network
“The final UDN cohort contains 465 patients representing 319 MONDO diseases and 378 unique causal genes.”
Rated minor: it would not touch a result — a value no result rests on, a p-value that stays on the same side of the significance level, or a conflict that is clerical, resolved elsewhere in the paper, or outside its own findings.
In the paper
§ Few shot learning for phenotype-driven diagnosis of patients with rare genetic diseases > Results > Overview of the undiagnosed diseases network patient cohort
“UDN patients have heterogeneous disease presentations: 378 unique genes and 299 unique diseases are represented in the cohort, and 48% of phenotype terms, 79% of genes, and 83% of diseases are represented in only a single patient (Fig. 1d). This reinforces the need for machine learning models that can learn from sparsely labeled datasets. 11.4% of patients have a diagnosis in common with at least one other patient.”
§ Few shot learning for phenotype-driven diagnosis of patients with rare genetic diseases > Results > Overview of the undiagnosed diseases network patient cohort
“We assemble a cohort of 465 patients in the Undiagnosed Diseases Network (UDN) with molecular diagnoses.”
Rated minor: it would not touch a result — a value no result rests on, a p-value that stays on the same side of the significance level, or a conflict that is clerical, resolved elsewhere in the paper, or outside its own findings.
In the paper
§ Few shot learning for phenotype-driven diagnosis of patients with rare genetic diseases > Patients in the Undiagnosed Diseases Network
“The EXPERT-CURATED and VARIANT-FILTERED candidate gene lists contain 244.3 and 13.3 genes on average, respectively (SD = 244.0 and SD = 8.0).”
§ Few shot learning for phenotype-driven diagnosis of patients with rare genetic diseases > Results > Overview of the undiagnosed diseases network patient cohort
“The two candidate gene lists contain 244.3 and 13.3 genes on average, respectively (SD = 244.0 and SD = 8.0; Fig. 1c).”
Citations
Cites
Richards, S., Aziz, N., Bale, S., Bick, D., Das, S., Gastier-Foster, J., Grody, W. W., Hegde, M., Lyon, E., Spector, E., et al. (2015) · Genetics in Medicine · doi:10.1038/gim.2015.30 · Cited by 32,594 · cited 1×
“TheEXPERT-CURATEDcandidategenelistfor eachpatientincludestheunionofboth(1)disease-associatedandother clinically-relevantgeneslistedonthepatient ’ s clinical sequencing reports (ACMG)guidelines 29,101,102 and(2)genes that were prioritized by UDN clinical teams who handled the patient ’ s case.”
PyTorch: An Imperative Style, High-Performance Deep Learning Library
Paszke, A., Gross, S., Massa, F., Lerer, A., Bradbury, J., Chanan, G., Killeen, T., Lin, Z., Gimelshein, N., Antiga, L., et al. (2019) · arXiv · doi:10.48550/arxiv.1912.01703 · Cited by 31,473
Dermatologist-level classification of skin cancer with deep neural networks
Esteva, A., Kuprel, B., Novoa, R. A., Ko, J., Swetter, S. M., Blau, H. M., & Thrun, S. (2017) · Nature · doi:10.1038/nature21056 · Cited by 14,362
Gephi: An Open Source Software for Exploring and Manipulating Networks
Bastian, M., Heymann, S., & Jacomy, M. (2009) · Proceedings of the International AAAI Conference on Web and Social Media · doi:10.1609/icwsm.v3i1.13937 · Cited by 11,675
High-performance medicine: the convergence of human and artificial intelligence
Topol, E. J. (2019) · Nature Medicine · doi:10.1038/s41591-018-0300-7 · Cited by 10,450
UMAP: Uniform Manifold Approximation and Projection
McInnes, L., Healy, J., Saul, N., & Großberger, L. (2018) · Journal of Open Source Software · doi:10.21105/joss.00861 · Cited by 9,826
Szklarczyk, D., Gable, A. L., Nastou, K. C., Lyon, D., Kirsch, R., Pyysalo, S., Doncheva, N. T., Legeay, M., Fang, T., Bork, P., et al. (2021) · Nucleic Acids Research · doi:10.1093/nar/gkaa1074 · Cited by 8,056 · cited 2×
“…hyperparameters: learning rate ∈ [0.00001,0.00005,0.0001,0.0005,0.001], λ ∈ [0.1,0.9], η ∈ [0.1,0.9], k most infrequently seen genes ∈ [64, 128, 192], number of transformer layers for the phenotype encoder ∈ [0, 3, 6], number of heads in the transformer layers ∈ [4, 8, 16], contribution…”
InfluentialGulshan, V., Peng, L., Coram, M., Stumpe, M. C., Wu, D., Narayanaswamy, A., Venugopalan, S., Widner, K., Madams, T., Cuadros, J., et al. (2016) · JAMA · doi:10.1001/jama.2016.17216 · Cited by 7,697
The Reactome pathway Knowledgebase
Fabregat, A., Sidiropoulos, K., Garapati, P., Gillespie, M., Hausmann, K., Haw, R., Jassal, B., Jupe, S., Korninger, F., McKay, S., et al. (2016) · Nucleic Acids Research · doi:10.1093/nar/gkv1351 · Cited by 5,605
Object Detection With Deep Learning: A Review
Zhao, Z.-Q., Zheng, P., Xu, S.-T., & Wu, X. (2019) · IEEE Transactions on Neural Networks and Learning Systems · doi:10.1109/tnnls.2018.2876865 · Cited by 5,545
Research works printed but not linked: 6
- Brody, S., Alon, U. & Yahav, E. How attentive are graph attention networks? ICLR (2022).
- Yang, B., Yih, W.-t., He, X., Gao, J. & Deng, L. Embedding entities and relations for learning and inference in knowledge bases. ICLR (2015).
- Goldberger, J., Hinton, G. E., Roweis, S. & Salakhutdinov, R. R. Neighbourhood components analysis. In Proc. 18th International Conference on Neural Information Processing Systems 513–520 (MIT Press, 2004).
- Ba, J. L., Kiros, J. R. & Hinton, G. E. Layer normalization. NeurIPS (2016).
- Loffe, S. & Szegedy, C. Batch normalization: Accelerating deep network training by reducing internal covariate shift. In International conference on machine learning, 448–456 (2015).
- Fey, M. & Lenssen, J. E. Fast graph representation learning with PyTorch Geometric. In ICLR Workshop on Representation Learning on Graphs and Manifolds (2019).
Cited by
A foundation model for clinician-centered drug repurposing
Huang, K., Chandak, P., Wang, Q., Havaldar, S., Vaid, A., Leskovec, J., Nadkarni, G. N., Glicksberg, B. S., Gehlenborg, N., & Zitnik, M. (2024) · Nature Medicine · doi:10.1038/s41591-024-03233-x · Cited by 249
“Addressing the needs of these diseases, many of which are complex, neglected or rare, is a clinical priority 2,49 .”
Self-citationBiomni: A general-purpose biomedical ai agent
Huang, K., Zhang, S., Wang, H., Qu, Y., Lu, Y., Roohani, Y., Li, R., Qiu, L., Li, G., Zhang, J., et al. (2025) · doi:10.1101/2025.05.30.656746 · Cited by 125
Rapid genomic sequencing for genetic disease diagnosis and therapy in intensive care units: a review
Kingsmore, S. F., Nofsinger, R. R., & Ellsworth, K. (2024) · npj Genomic Medicine · doi:10.1038/s41525-024-00404-0 · Cited by 104
An agentic system for rare disease diagnosis with traceable reasoning
Zhao, W., Wu, C., Fan, Y., Qiu, P., Zhang, X., Sun, Y., Zhou, X., Zhang, S., Peng, Y., Wang, Y., et al. (2026) · Nature · doi:10.1038/s41586-025-10097-9 · Cited by 62
Traylor, D., Kern, K. V., Anderson, E. E., & Henderson, R. (2025) · Cureus · doi:10.7759/cureus.76825 · Cited by 48
Metadata
- Authors
- Alsentzer, E., Li, M. M., Kobren, S. N., Noori, A., Undiagnosed Diseases Network, Kohane, I. S., & Zitnik, M.
- Affiliations
- Harvard University (US)
- Broad Institute (US)
- Venue
- npj Digital Medicine
- Published
- 2025-06-20
- Volume
- 8
- Issue
- 1
- Pages
- 380-380
- ISSN
- 2398-6352 (ISSN-L)
- PMC
- PMC12181314
- OpenAlex
- W4411485357
- Cited by
- 54
- References
- 127
- Citations by year
- 2026: 37 · 2025: 12
- FWCI
- 28.59
- Type
- journal article
- Language
- en
- Open access
- gold
- doi.org
- Licence
- cc-by-nc-nd
- creativecommons.org/licenses/by-nc-nd/4.0 (tdm)
- creativecommons.org/licenses/by-nc-nd/4.0 (vor)
- Locations
- npj Digital Medicine (publishedVersion, cc-by-nc-nd)
- medRxiv (acceptedVersion, cc-by)
- PubMed Central (publishedVersion, cc-by-nc-nd)
- PubMed (publishedVersion)
- DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion, cc-by-sa)
- Funders
- National Science Foundation (2030459, 2033384)
- AstraZeneca
- Microsoft Research
- Wellcome Trust (WT223718/Z/21/Z)
- European Bioinformatics Institute
- National Institutes of Health (U01TR001395, -0001, U01TR002471)
- U.S. Air Force (FA8702‐15‐D‐0001)
- Harvard Data Science Initiative, Harvard University
- National Human Genome Research Institute (U01HG010233, U01HG007709, U01HG007943, U01HG007703, U01HG007530, U01HG010230, U01HG007674, U01HG010217, U01HG007690, U01HG007942, U01HG007708, U01HG010219, U01HG010215, U01HG007672, T32HG002295)
- National Institute of Neurological Disorders and Stroke (U54NS093793, U54NS108251)
- National Center for Advancing Translational Sciences
- Common Fund (U01HG007530, U01TR001395, U01HG007709, U01HG007690, U01HG007672, U01HG007942, U01HG007708, U01HG007943, U54NS093793, U01 HG007703)
- Topics
- Life Sciences › Biochemistry, Genetics and Molecular Biology › Genetics › Genomics and Rare Diseases
- Physical Sciences › Computer Science › Artificial Intelligence › Machine Learning in Healthcare
- Health Sciences › Medicine › Health Informatics › Artificial Intelligence in Healthcare and Education
- Keywords
- rare genetic disorders, few-shot learning, rare disease diagnosis, knowledge graphs, patient retrieval, Undiagnosed Diseases Network, developmental disorders, phenotypic heterogeneity
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.
The claim verdicts this report counts are experimental: their accuracy is not yet measured.
A language model drew the claims from the paper on 2026-10-10, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).
A second model labelled each claim on 2026-10-10 (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
How to cite
Cite CoChat for the checks on this page, and the paper for its own findings.
- This page
- CoChat. (2026). Citation check: Few shot learning for phenotype-driven diagnosis of patients with rare genetic diseases. https://app.cochat.ai/papers/10.1038/s41746-025-01749-1
- The paper
- Alsentzer, E., Li, M. M., Kobren, S. N., Noori, A., Undiagnosed Diseases Network, Kohane, I. S., & Zitnik, M. (2025). Few shot learning for phenotype-driven diagnosis of patients with rare genetic diseases. npj Digital Medicine. https://doi.org/10.1038/s41746-025-01749-1
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.