Structured multi-view relational learning for drug response prediction.

MulDR, a structured multi-view relational learning framework for drug response prediction, reduces root mean square error by 6.07% versus the strongest baseline under the same evaluation protocol.

• Why it matters: Drug response prediction needs models that capture how molecular drug structure interacts with heterogeneous cellular states, but existing multi-omics methods can blur view-specific cell-line relations.
• What they did: MulDR builds separate relation structures for pathway activity, gene expression, mutation, and DNA methylation, then learns view-specific cell representations and fuses them using drug-conditioned motif-level, geometry-aware molecular graphs, and Morgan fingerprints.
• The result: Across extensive experiments, MulDR outperforms representative and recent baselines, generalizes in cold-start settings for unseen drugs and cell lines, and ablations show that removing view-specific relational modeling or drug-conditioned fusion degrades performance.

AI summary of the abstract · may contain errors · not medical advice · read the paper.