A foundation model for clinician-centered drug repurposing

In the paper

“This evaluation method aligns with the approach predominantly used in the literature12,14–16,18–20.”

What the sources say

• Supported

  The Methods section, “Evaluating predictions of what drugs will treat a disease” (page not provided), describes the paper’s drug-ranking evaluation and explicitly says its assumption about drugs not approved for a disease is consistent with prior literature. That supports the cited sentence’s claim of alignment with an established evaluation approach. (LLM, conf=0.84)

  cites [12] Ruiz, C., Zitnik, M. & Leskovec, J. Identification of disease treatment mechanisms through the multiscale interactome. Nat. Commun. 12, 1–15 (2021). source
• ?Could not verify

  The Network pharmacology section discusses network-based drug discovery and drug-target networks, but does not describe a standard benchmarking strategy for drug-repurposing AI models or say which evaluation approach predominates in the literature. A search for benchmarking, validation, and predictive-performance approaches found no passage addressing that claim. (LLM, conf=0.76)

  cites [14] Baraba´si, A.-L., Gulbahce, N. & Loscalzo, J. Network medicine: a network-based approach to human disease. Nat. Rev. Genet. 12, 56–68 (2011). sourceExperimental — accuracy not yet measured
• ?Could not verify

  The ‘Graph representation learning for therapeutics’ section discusses methods for predicting drug–disease associations and ranking repurposing candidates, but it does not describe a standard benchmarking strategy for evaluating those models. Searches for evaluation protocols, benchmarks, and prevalent practice found no passage establishing that the citing paper’s unspecified evaluation method is predominantly used in the literature. The retrieved passages have no page numbers in our copy. (LLM, conf=0.72)

  cites [15] Li, M. M., Huang, K. & Zitnik, M. Graph representation learning in biomedicine and healthcare. Nat. Biomed. Eng. 6, 1353–1369 (2022). sourceExperimental — accuracy not yet measured
• Partly supported

  The Abstract says that repurposing algorithms typically validate only a small subset of top-ranked candidates, supporting the claim that a common evaluation practice exists. But it argues that reliable comparison requires testing a wider, rank-independent set of drugs; it does not establish that the citing paper’s unspecified evaluation method aligns with the predominant practice. (LLM, conf=0.78)

  cites [16] Gysi, D. M. et al. Network medicine framework for identifying drug-repurposing opportunities for Covid-19. Proc. Natl Acad. Sci. USA 118, e2025581118 (2021). sourceExperimental — accuracy not yet measured
• Partly supported

  The Experimental setup section (page unavailable) describes evaluating the model with training, validation, and test splits and reporting AUROC, AUPRC, and AP@50. This supports the paper as an example of an AI-model evaluation approach, but it does not establish that the citing paper’s unspecified method is the approach predominantly used in drug-repurposing literature. (LLM, conf=0.76)

  cites [18] Zitnik, M., Agrawal, M. & Leskovec, J. Modeling polypharmacy side effects with graph convolutional networks. Bioinformatics 34, i457–i466 (2018). sourceExperimental — accuracy not yet measured
• Partly supported

  The Methods section (‘Area under ROC curve and optimal proximity cutoff analysis’; page unavailable) describes evaluating drug–disease predictions with ROC-AUC, using known associations as positives and unknown associations as proxy negatives. This establishes that the paper uses an evaluation approach of that kind, but it does not establish that the approach is predominant in the literature. (LLM, conf=0.83)

  cites [19] Guney, E., Menche, J., Vidal, M. & Bara´basi, A.-L. Network-based in silico drug efficacy screening. Nat. Commun. 7, 1–13 (2016). sourceExperimental — accuracy not yet measured
• Not checked yetcites [20] Cheng, F., Kova´cs, I. A. & Baraba´si, A.-L. Network-based prediction of drug combinations. Nat. Commun. 10, 1–11 (2019). source

In the paper

“Here we introduce TxGNN, a graph foundation model for zero-shot drug repurposing, identifying therapeutic candidates even for diseases with limited treatment options or no existing drugs.”

• Also stated in: § Main · plain

  “In the present study, we introduced TxGNN, a graph foundation model for multi-disease, zero-shot drug repurposing that predicts drug-repurposing candidates across 17,080 diseases, including diseases without treatments (Fig. 1c).”

In the paper

“Foundation models like TxGNN are transforming deep learning: instead of training disease-specific models for every disease, TxGNN is a single pretrained model that adapts across many diseases.”

In the paper

“When benchmarked against 8 methods, TxGNN improves prediction accuracy for indications by 49.2% and contraindications by 35.1% under stringent zero-shot evaluation.”

In the paper

“Human evaluation of TxGNN’s Explainer showed that TxGNN’s predictions and explanations perform encouragingly on multiple axes of performance beyond accuracy.”

• Also stated in: § Main · plain

  “Our human evaluation of TxGNN’s Explainer showed that TxGNN’s explanations perform encouragingly on multiple axes of performance, including accuracy, trust, usefulness and time efficiency.”

In the paper

“BioBERT performed best for indication prediction in seven of nine disease areas, whereas RGCN was the best baseline for contraindications in eight of nine.”

In the paper

“Although established methods achieved satisfactory results in conventional drug-repurposing evaluations, they often fell short in challenging settings.”

In the paper

“TxGNN achieved relative AUPRC gains of 0.5–59.3% (average 25.72%) across 9 disease areas for indications and 11.8–35.6% (average 18.67%) for contraindications.”

In the paper

“However, TxGNN Explainer indicated that zolpidem’s action on GABRG2 might reduce autism susceptibility and improve prefrontal cortex function.”

In the paper

“Tretinoin, transported by albumin (ALB) and targeting ALDH1A2, helps mitigate collagen loss and inflammation, as highlighted in TxGNN’s predictive rationale (Fig. 5c), indicating that TxGNN’s predictive rationale is aligned with medical reasoning.”

In the paper

“Focusing on the most predictive relationships (that is, edges with importance scores >0.5, representing an average of 14.9% of edges from the KG), the model’s performance showed a slight reduction from AUPRC = 0.890 (s.d. 0.006) to AUPRC = 0.886 (s.d. 0.005), indicating that TxGNN’s Explainer effectively identifies key associations34 and explanations accurately reflect TxGNN’s internal reasoning.”

What the sources say

• Not checked yetcites [34] Agarwal, C., Queen, O., Lakkaraju, H. & Zitnik, M. Evaluating explainability for graph neural networks. Sci. Data 10, 144 (2023). source

In the paper

“Our experimental results in this setting concur, with three of eight existing methods achieving AUPRC > 0.80 and HAN as the best at 0.873 AUPRC.”

In the paper

“Although the average log(OR) stands at 1.09, the top therapeutic candidate predicted by TxGNN had a log(OR) of 2.26, approaching the average log(OR) of 2.92 for FDA-approved indications, indicating the enrichment of off-label drug prescriptions among TxGNN’s top-ranked predictions.”

In the paper

“The medical KG covers 17,080 diseases, 92% lacking FDA-approved drugs, covering rare and less-understood complex diseases.”

In the paper

“TxGNN substantially improves over the next best method in predicting indications (19.0% AUPRC gain) and contraindications (23.9% AUPRC gain).”

In the paper

“TxGNN consistently ranked newly introduced drugs favorably and, in two instances, placed the newly approved medications within the top 5% of predicted drugs.”

In the paper

“However, TxGNN outperformed all methods across all nine disease areas for both tasks, demonstrating its broad generalizability and accuracy in zero-shot drug repurposing.”

In the paper

“As most diseases do not share underlying pathology, they have low similarity scores.”

In the paper

“Finally, we used parameter search and found optimal λ = 0.7.”

In the paper

“Drug repurposing leverages existing safety and efficacy data of approved drugs, allowing for faster translation to the clinic and reduced development costs compared to developing new drugs from scratch4 (Fig. 1a).”

What the sources say

• Not checked yetcites [4] Pushpakom, S. et al. Drug repurposing: progress, challenges and recommendations. Nat. Rev. Drug Discov. 18, 41–58 (2019). source

In the paper

“There is a pressing need to develop therapies for many diseases that currently lack treatments1,2.”

What the sources say

• Not checked yetcites [1] Feigin, V. L. et al. Burden of neurological disorders across the us from 1990-2017: a global burden of disease study. JAMA Neurol. 78, 165–176 (2021). source
• Not checked yetcites [2] O’Connell, D. Neglected diseases. Nature 449, 157–157 (2007). source

In the paper

“Of over 7,000 rare diseases worldwide, only 5–7% have Food and Drug Administration (FDA)-approved drugs3.”

What the sources say

• Not checked yetcites [3] Rare Disease Day 2021. US Food and Drug Administration fda.gov/news-events/fda-voices/rare-disease-day-2021-fda-shows-sustained-support-rare-disease-product-development-during-public (2023).

In the paper

“Approximately 30% of FDA-approved drugs are issued at least one new post-approval indication and many drugs have accrued over ten indications over the years6.”

What the sources say

• Not checked yetcites [6] Sahragardjoonegani, B., Beall, R. F., Kesselheim, A. S. & Hollis, A. Repurposing existing drugs for new uses: a cohort study of the frequency of FDA-granted new indication exclusivities since 1997. J. Pharm. Policy Pract. 14, 3 (2021). source

In the paper

“Moreover, around 95% of rare diseases have no FDA-approved drugs and up to 85% of rare diseases do not have even one drug developed that would show promise in rare disease treatment, diagnosis or prevention21.”

What the sources say

• Not checked yetcites [21] Fermaglich, L. J. & Miller, K. L. A comprehensive study of the rare diseases and conditions targeted by orphan drug designations and approvals over the forty years of the orphan drug act. Orphanet J. Rare Dis. 18, 1–8 (2023). source

In the paper

“Originally developed to help with morning sickness during pregnancy, thalidomide was repurposed in 1964 for an autoimmune complication of leprosy and again in 2006 for multiple myeloma8.”

What the sources say

• Not checked yetcites [8] Jourdan, J.-P., Bureau, R., Rochais, C. & Dallemagne, P. Drug repositioning: a brief overview. J. Pharm. Pharmacol. 72, 1145–1151 (2020). source

In the paper

“Machine learning has been used to analyze high-throughput molecular interactomes to unravel genetic architecture perturbed in disease12,13 and help design therapies to target them14.”

What the sources say

• Not checked yetcites [12] Ruiz, C., Zitnik, M. & Leskovec, J. Identification of disease treatment mechanisms through the multiscale interactome. Nat. Commun. 12, 1–15 (2021). source
• Not checked yetcites [13] Goh, K.-I. et al. The human disease network. Proc. Natl Acad. Sci. USA 104, 8685–8690 (2007). source
• Not checked yetcites [14] Baraba´si, A.-L., Gulbahce, N. & Loscalzo, J. Network medicine: a network-based approach to human disease. Nat. Rev. Genet. 12, 56–68 (2011). source

In the paper

“These strategies identify therapeutic candidates based on their impact on cell signaling, gene expression and disease phenotypes5,10–12.”

What the sources say

• Not checked yetcites [5] Abdelsayed, M., Kort, E. J., Jovinge, S. & Mercola, M. Repurposing drugs to treat cardiovascular disease in the era of precision medicine. Nat. Rev. Cardiol. 19, 751–764 (2022). source
• Not checked yetcites [10] Menche, J. et al. Uncovering disease-disease relationships through the incomplete interactome. Science 347, 1257601 (2015). source
• Not checked yetcites [11] Zitnik, M. et al. Evolution of resilience in protein interactomes across the tree of life. Proc. Natl Acad. Sci. USA 116, 4426–4433 (2019). source
• Not checked yetcites [12] Ruiz, C., Zitnik, M. & Leskovec, J. Identification of disease treatment mechanisms through the multiscale interactome. Nat. Commun. 12, 1–15 (2021). source

In the paper

“However, Duran et al.30 reasoned that these models make predictions for a disease by retrieving drugs from the dataset that appear similar to existing treatments.”

What the sources say

• Not checked yetcites Duran et al.30 Duran-Frigola, M. et al. Extending the small-molecule similarity principle to all levels of biology with the chemical checker. Nat. Biotechnol. 38, 1087–1096 (2020). source

In the paper

“AVPR2 and NPR1 genes play pivotal roles in regulating fluid and electrolyte balance via complementary but distinct pathways. AVPR2 contributes to water retention and urine concentration, whereas NPR1 facilitates vasodilatation, lowers blood pressure and enhances water excretion47.”

What the sources say

• Not checked yetcites [47] Wishart, D. S. et al. DrugBank 5.0: a major update to the DrugBank database for 2018. Nucleic Acids Res. 46, D1074–D1082 (2018). source

In the paper

“Furthermore, some subtypes of Ehlers–Danlos syndrome have been associated with a pathogenic mutation in the ALB gene in ref. 42 and weakly linked to ALDH1AI in ref. 43.”

What the sources say

• Not checked yetcites [42] Landrum, M. J. et al. Clinvar: improvements to accessing data. Nucleic Acids Res. 48, D835–D844 (2020). source
• Not checked yetcites [43] Javed, S. et al. ALDH1 & CD133 in invasive cervical carcinoma & their association with the outcome of chemoradiation therapy. Indian J. Med. Res. 154, 367 (2021). source

In the paper

“This phenomenon is known as shortcut learning31 and underlies many deep-learning failures32.”

What the sources say

• Not checked yetcites [31] Bickel, S., Brückner, M. & Scheffer, T. Discriminative learning under covariate shift. J. Mach. Learn. Res. 10, 2137–2155 (2009).
• Not checked yetcites [32] Niven, T. & Kao, H.-Y. Probing neural network comprehension of natural language arguments. In Proc. of the 57th Annual Meeting of the Association for Computational Linguistics 4658–4664 (ACL, 2019). source

In the paper

“the literature indicates that it may effectively eliminate hepatic iron48.”

What the sources say

• Not checked yetcites [48] Seetharaman, J. & Sarma, M. S. Chelation therapy in liver diseases of childhood: current status and response. World J. Hepatol. 13, 1552 (2021). source

In the paper

“Surprisingly, zolpidem has shown stimulative effects in neurological conditions, temporarily awakening underactive neurons, suggesting a potential therapeutic use for neurodevelopmental disorders39.”

What the sources say

• Not checked yetcites [39] Bomalaski, M. N., Claflin, E. S., Townsend, W. & Peterson, M. D. Zolpidem for the treatment of neurologic disorders: a systematic review. JAMA Neurol. 74, 1130–1139 (2017). source

In the paper

“Shortcut decision rules tend to perform well on standard benchmarks but fail on challenging conditions33, such as predicting drugs for rare diseases with no treatment options or subtypes of complex diseases with distinct disease mechanisms.”

What the sources say

• Not checked yetcites [33] Geirhos, R. et al. Shortcut learning in deep neural networks. Nat. Mach. Intell. 2, 665–673 (2020). source

In the paper

“Amyl nitrite, which targets the NPR1 gene, emerges as a potential therapeutic option for NSIAD, confirming the consistency of TxGNN’s explanations with medical evidence.”

In the paper

“It is important to note that the weights in the decoder, specifically for DistMult, wr, are reinitialized before fine-tuning to mitigate the risk of negative knowledge transfer.”

In the paper

“We have complied with all relevant ethical regulations.”

In the paper

“The variant with the highest validation performance was selected for test set evaluation and downstream analyses.”

In the paper

“TxGNN streamlines drug-repurposing prediction, especially when the limited availability of disease-specific datasets hinders drug development.”

In the paper

“Despite this, rare diseases often offer extensive opportunities for therapeutic advancements3.”

What the sources say

• Not checked yetcites [3] Rare Disease Day 2021. US Food and Drug Administration fda.gov/news-events/fda-voices/rare-disease-day-2021-fda-shows-sustained-support-rare-disease-product-development-during-public (2023).

In the paper

“Our pilot human evaluation engaged a small number (n = 12) of clinicians and scientists and prioritized an in-depth analysis with a small but qualified group of human experts over a broader study with a larger, potentially less specialized, participant pool.”

In the paper

“TxGNN is a unified model for predicting indications and contraindications across 17,080 diseases, suitable for early drug repurposing beyond single therapeutic areas.”

In the paper

“Informed consent was waived for analyzing EMRs retrospectively.”

In the paper

“Appropriate provisions were taken to ensure the safety and well-being of all participants involved.”

In the paper

“Despite the promising performance of TxGNN’s predictions on medical records, unaccounted confounders and selection biases might have limited the ability to draw conclusions from the calculated drug enrichment scores.”

In the paper

“Multi-hop interpretable predictive rationales can enhance transparent use of TxGNN, fostering trust and aiding human experts.”

In the paper

“TxGNN generalizes to diseases with few treatment options by modeling latent relationships between diseases and performing zero-shot inference for diseases that the model never encountered during training.”

In the paper

“Use of a learnable attention mechanism is ineffective, because it overvalues the original embeddings for well-represented diseases, neglecting the auxiliary embedding.”

In the paper

“Special consideration is needed because, for well-represented diseases in the KG, the disease–disease metric learning embedding might be unnecessary and could bias the disease embedding.”

In the paper

“The sequence in which predictions were presented, along with the conditions (TxGNN Explorer or the baseline approach), were randomized and counterbalanced across participants and tasks.”

In the paper

“These objectives inherently entail a tradeoff: increasing the extent of masking tends to enlarge the discrepancy between the updated and original predictions.”

In the paper

“The findings from this comparison motivated the adoption of path-based explanations, which were preferred based on user feedback.”

What the sources say

• Not checked yetcites [25] Wang, Q., Huang, K., Chandak, P., Zitnik, M. & Gehlenborg, N. Extending the nested model for user-centric XAI: a design study on gnn-based drug repurposing. IEEE Trans. Vis. Comput. Graph. 29, 1266–1276 (2023). source

In the paper

“Given the insufficiency of disease representations produced solely by GNNs, these representations are not ideal for direct similarity computations.”

In the paper

“Using a model to extract predictive information from similar but better-represented diseases in the KG, the target disease embedding can be enriched, improving model performance for the disease.”

In the paper

“All internal EMRs were deidentified before computational analysis and model development.”

In the paper

“Our research did not result in stigmatization, incrimination, discrimination or personal risk to participants.”

In the paper

“TxGNN’s zero-shot drug-repurposing capability allows the model to predict drugs for diseases with limited treatment options and scarce information.”

In the paper

“For instance, the molecular basis of many rare diseases remains poorly understood21.”

What the sources say

• Not checked yetcites [21] Fermaglich, L. J. & Miller, K. L. A comprehensive study of the rare diseases and conditions targeted by orphan drug designations and approvals over the forty years of the orphan drug act. Orphanet J. Rare Dis. 18, 1–8 (2023). source

In the paper

“After the training process is complete, edges (i, j, r) for which = 0 can be removed. The remaining edges serve as explanations for the model’s predictions.”

In the paper

“The Mount Sinai Institutional Review Board approved the study, ensuring that all clinical data were deidentified.”

In the paper

“Previous research has emphasized KG completion, optimizing models across the entire spectrum of relationships within a KG58.”

What the sources say

• Not checked yetcites [58] Lin, Y., Liu, Z., Sun, M., Liu, Y. & Zhu, X. Learning entity and relation embeddings for knowledge graph completion. In Proc. of the 29th AAAI Conference on Artificial Intelligence Vol. 29, 2181–2187 (2015). source

In the paper

“A higher degree indicated that the disease was better represented in the knowledge and had a denser local network neighborhood, suggesting a reduced reliance on the disease–disease metric learning embedding and vice versa.”

In the paper

“Predicted drugs align with off-label prescription rates in EMRs and match with the medical consensus of human experts.”

In the paper

“Two-sided Bonferroni’s correction was applied to the P values using the statsmodels Python library’s multi-test function, identifying statistically significant drug–disease pairs as those with P < 0.005.”

In the paper

“The Institutional Review Board at Mount Sinai, New York City, USA, approved the retrospective analysis of internal EMRs.”

In the paper

“Predictions were generated for 1,363 diseases, identified by training a KG on 5% of randomly selected drug–disease pairs, serving as a validation set for early stopping.”