Graph Network-Based Analysis of Disease-Gene-Drug Associations: Zero-Shot Disease-Drug Prediction and Analysis Strategies
Bibliographic record
Abstract
Abstract Existing drug repurposing methods have key limitations, primarily stemming from their reliance on known direct associations between diseases and drugs for supervised learning, as well as the need for large amounts of prior disease or drug information or feature data. In practice, many disease-drug connections remain unknown, and prior information is often complex and difficult to acquire and organize, limiting the applicability of these models. Furthermore, these models generally lack interpretability, making it difficult for experts to assess the reliability of predictions based solely on standard metrics, which raises doubts about the trustworthiness of their results. To address these challenges, we propose ZS-GNT, an innovative new workflow for zero-shot drug repurposing that leverages a novel and ingenious graph data meta-path linking scheme, which does not require any known disease-drug associations or their prior features. This approach is implemented using the Graph Neural Transformer (GNT) algorithm. The method infers disease-drug relationships indirectly through gene action, utilizing disease-gene associations and gene-drug interactions. It also generates a top drug-top gene linkage map, providing clinicians with a visual tool to assess the plausibility of suggested drugs before advancing to clinical trials. Experimental results show that, under the same linking scheme, the GNT algorithm achieved interaction link prediction accuracies of 95.86%, 99.28%, and 99.54% for three diseases, surpassing four other baseline methods. In a test involving a random selection of 100 diseases for drug discovery, among the top 5 recommended drugs from the candidates identified by ZS-GNT from a pool of 33,251 total drugs, the validation rate reached 47.05%, demonstrating the model’s effectiveness in drug discovery.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.002 | 0.005 |
| Meta-epidemiology (narrow) | 0.001 | 0.001 |
| Meta-epidemiology (broad) | 0.001 | 0.002 |
| Bibliometrics | 0.004 | 0.002 |
| Science and technology studies | 0.001 | 0.001 |
| Scholarly communication | 0.001 | 0.002 |
| Open science | 0.002 | 0.001 |
| Research integrity | 0.001 | 0.001 |
| Insufficient payload (model declined to judge) | 0.002 | 0.001 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".