Benchmark testing whether frontier models can recover a biological relatedness signal that a dominant confounder hides. 58 questions over protein, nucleotide, and scRNA-seq fixtures with deterministic ground truth.
Neutral leaderboard for protein conformational ensemble generators, scored on a fixed metric panel over the ATLAS molecular dynamics set.
PyTorch library for parametric dimensionality reduction: t-SNE, UMAP, PaCMAP, TriMap, and CEBRA under a single API.
Python/TensorFlow/Keras implementation of parametric t-SNE for dimensionality reduction.
Protein structure prediction, enabling accurate 3D modeling of protein conformations.
Molecular docking using diffusion generative models for predicting ligand binding poses.
Lung cancer risk prediction from low-dose CT scans (MIT Barzilay group).
Breast cancer risk prediction from mammography (MIT Barzilay group).
Contributions to the scientific Python ecosystem.