Writing & Research Notes

Structure-Based Prediction of Autophagy Modulators: From Transcriptome Collapse to Binary Classifiers

A rigorous benchmarking study of machine learning models for autophagy prediction using LINCS L1000 and HAMDB datasets. Explores why structure-to-transcriptome regression collapses on canonical controls, why generic stress programs confound signature matching, and why Morgan fingerprint binary classifiers establish a clean 0.858 AUROC baseline.