Researchtransformersdrug repurposingembeddingsinterpretability
FKSUDDAPre Predicts Drug–Disease Associations With High Accuracy
8.2
Zuo et al. publish FKSUDDAPre on February 5, 2026, a framework that predicts drug–disease associations using Mol2vec and K-BERT for drugs and MeSH with DeepWalk for diseases. The system uses AMDKSU resampling, F-test feature selection, and an XGBoost/Decision Tree/Random Forest/HyperFast ensemble, achieves average AUC 0.9725 (≈3.88% improvement), and validates many top candidates for Alzheimer’s and Parkinson’s disease.
Key Points
- 1Combines Mol2vec and K-BERT plus MeSH–DeepWalk to fuse drug and disease representation features.
- 2Implements AMDKSU resampling and F-test selection to address imbalance and reduce feature dimensionality.
- 3Achieves average AUC 0.9725 and validates top candidates for Alzheimer’s and Parkinson’s disease.
Scoring Rationale
Peer-reviewed validation, strong performance, and usable code increase score; novelty is moderate compared to existing DDA methods.
Sources
Public references used for this report.
Practice interview problems based on real data
1,625 SQL & Python problems across 15 industry datasets — the exact type of data you work with.
Try 250 free problems
