- Genomics and Phylogenetic Studies
- Machine Learning in Bioinformatics
- Protein Structure and Dynamics
- Evolutionary Algorithms and Applications
- Gene expression and cancer classification
University of California, Berkeley
2022-2023
Computational genomics increasingly relies on machine learning methods for genome interpretation, and the recent adoption of neural sequence-to-function models highlights need rigorous model specification controlled evaluation, problems familiar to other fields AI. Research strategies that have greatly benefited -- including benchmarking, auditing, algorithmic fairness --- are also needed advance field genomic AI facilitate development. Here we propose a benchmark, GUANinE, evaluating...
Protein language models have enabled breakthrough approaches to protein structure prediction, function annotation, and drug discovery. A primary limitation the widespread adoption of these powerful is high computational cost associated with training inference models, especially at longer sequence lengths. We present architecture, microarchitecture, hardware implementation a design discovery accelerator, ProSE (Protein Systolic Engine). has collection custom heterogeneous systolic arrays...
The nascent field of genomic AI is rapidly expanding with new models, benchmarks, and findings. As the diversifies, there an increased need for a common set measurement tools perspectives to standardize model evaluation. Here, we present statistically grounded framework performance evaluation, visualization, interpretation using prominent Enformer as case study. has been used range applications from mechanism discovery variant effect prediction, but what makes it better or worse than...