Daniel Himmelstein

This paper develops a statistical learning framework to predict clinical trial success for target-disease pairs using longitudinal evidence, outperforming existing scores and identifying high-potential targets.

The paper introduces a network permutation framework to assess how node degree influences edge prediction in biomedical networks and provides an edge prior as a baseline for performance evaluation.

The paper introduces hetnet connectivity search to find meaningful relationships between biomedical entities in heterogeneous networks without requiring supervised training data.

The paper introduces the Bioregistry, an open metaregistry that integrates and expands existing biomedical entity registries to improve interoperability and data reuse.

The paper improves knowledge graph expansion by reusing label functions for multiple edge types in biomedical abstracts.

The paper analyzes disparities in recognition of computational biology researchers based on gender, name origin, country, and race/ethnicity.

The paper evaluates multiple compression methods and dimensionalities on gene expression data, finding no single best approach but showing benefits from ensemble methods across different dimensions.

Deep learning rivals or beats baselines across imaging, health records, genomics and drug discovery but is not yet transformative; scarce labels, privacy and opacity remain hurdles.

This paper uses a biomedical knowledge network to predict drug-disease treatment pairs and prioritize drug repurposing candidates.