1
ProteinGym: Large-Scale Benchmarks for Protein Fitness Prediction and Design
Large-scale benchmarks to assess models for protein fitness prediction and design.
ProteinNPT: Improving Protein Property Prediction and Design with Non-Parametric Transformers
A conditional semi-supervised pseudo-generative model for fitness prediction and design.
DiscoBAX: Discovery of Optimal Intervention Sets in Genomic Experiment Design
We introduce a sample-efficient method for discovering optimal sets that are both diverse and optimize the function of interest.
TranceptEVE: Combining Family-specific and Family-agnostic Models of Protein Sequences for Improved Fitness Prediction
TranceptEVE combines family-specific and family-agnostic models to achieve SOTA performance on protein fitness prediction and human …
Tranception: Protein Fitness Prediction with Autoregressive Transformers and Inference-time Retrieval
Tranception combines large autoregressive transformers with inference-time retrieval to achieve SOTA performance on protein fitness …
RITA: a Study on Scaling Up Generative Protein Sequence Models
We introduce RITA, a suite of autoregressive generative models for protein sequences, with up to 1.2 billion parameters
Improving black-box optimization in VAE latent space using decoder uncertainty
We use the epistemic uncertainty of the VAE decoder to guide the optimization of properties of high-dimensional structured objects …
Viral Evolution and Antibody Escape Mutations using Deep Generative Models
We leverage deep generative models of evolutionary sequences to predict viral escape mutations.
Principled Uncertainty Estimation for High Dimensional Data
We introduce an importance sampling-based estimator to estimate the epistemic uncertainty of deep learning models for high-dimensional …