Solip Park

This paper introduces SubNetDL, a deep learning model that uses subclonal mutation profiles and protein-protein interaction networks to predict cancer therapy response, achieving strong performance across multiple cancer types and treatment modalities.

The paper analyzes mutations and copy-number alterations in 25,000 tumor samples to identify state-specific interactions in cancer genes, revealing dynamic progression mechanisms.

The paper identifies new cancer predisposing genes by analyzing rare germline variants in Mendelian disease genes across cancer cohorts.