Archival

Proteins are essential for almost all biological processes and derive their diverse functions from complex 3D structures, which are in …

Matching objectives underpin the success of modern generative models and rely on constructing conditional paths that transform a source …

Efficiently generating statistically independent samples from an unnormalized probability distribution, such as equilibrium samples of …

The dynamical formulation of the optimal transport can be extended through various choices of the underlying geometry (kinetic energy), …

We present FoldFlow a novel flow matching model for protein design. We present theory and practical tricks for flow models over SE(3)^N. Empirically, we validate these models on protein backbone generation of up to 300 amino acids leading to high-quality designable, diverse, and novel samples.

We present simulation-free score and flow matching ([SF]2M), a simulation-free objective for inferring stochastic dynamics given …

Trellis is a tree-based earth mover’s distance method for understanding estimating treatment effects from single cell data. In this work we apply it to colorectal cancer PDOS and investigate the chemoprotection induced by cancer-associated fibroblasts.