Knowledge graph embeddings in the biomedical domain: are they useful? A look at link prediction, rule learning, and downstream polypharmacy tasks
Why this work is in the frame
A frame that forgets how it found something cannot be audited. These are the routes that admitted this work.
Bibliographic record
Abstract
Summary: Knowledge graphs (KGs) are powerful tools for representing and organizing complex biomedical data. They empower researchers, physicians, and scientists by facilitating rapid access to biomedical information, enabling the discernment of patterns or insights, and fostering the formulation of decisions and the generation of novel knowledge. To automate these activities, several KG embedding algorithms have been proposed to learn from and complete KGs. However, the efficacy of these embedding algorithms appears limited when applied to biomedical KGs, prompting questions about whether they can be useful in this field. To that end, we explore several widely used KG embedding models and evaluate their performance and applications using a recent biomedical KG, BioKG. We also demonstrate that by using recent best practices for training KG embeddings, it is possible to improve performance over BioKG. Additionally, we address interpretability concerns that naturally arise with such machine learning methods. In particular, we examine rule-based methods that aim to address these concerns by making interpretable predictions using learned rules, achieving comparable performance. Finally, we discuss a realistic use case where a pretrained BioKG embedding is further trained for a specific task, in this case, four polypharmacy scenarios where the goal is to predict missing links or entities in another downstream KGs in four polypharmacy scenarios. We conclude that in the right scenarios, biomedical KG embeddings can be effective and useful. Availability and implementation: Our code and data is available at https://github.com/aryopg/biokge.
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.
Full frame distilled prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.
Codex and Gemma teacher scores by category
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.001 | 0.000 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.000 |
| Bibliometrics | 0.001 | 0.002 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.000 | 0.001 |
| Open science | 0.002 | 0.001 |
| Research integrity | 0.000 | 0.002 |
| Insufficient payload (model declined to judge) | 0.000 | 0.000 |
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 it