Amaro N Taylor-Weiner

DINOcell: Learning Generalizable Perturbation Effects through Self-Distillation

DINOcell predicts cellular responses to perturbations using self-distillation in single-cell transcriptomics, outperforming baselines for single and combination gene effects.

Spatial mapping of immunosuppressive cancer-associated fibroblast gene signatures in H&E-stained images using additive multiple instance learning

This paper introduces aMIL, a machine learning method that predicts CAF gene signatures from H&E images, revealing spatial patterns linked to immunosuppression in tumors.

Dense, high-resolution mapping of cells and tissues from pathology images for the interpretable prediction of molecular phenotypes in cancer

This paper develops a method to predict molecular phenotypes from cancer pathology images using interpretable features, showing correlations with tumor microenvironment markers and predictive power for molecular signatures.

Discovery of cancer driver genes based on nucleotide context

This study identifies novel cancer driver genes by analyzing mutation nucleotide contexts in whole-exome sequencing data from 12,004 tumors.

Scaling computational genomics to millions of individuals with GPUs

This paper demonstrates that GPU-accelerated machine learning libraries significantly speed up and reduce the cost of computational genomics tasks like QTL mapping and Bayesian non-negative matrix factorization.

Passenger Hotspot Mutations in Cancer

This paper shows that many cancer hotspots are passengers, not drivers, and introduces a new model to better identify true driver mutations.