Clara Fannjiang

This paper introduces EVFI and DeepEVFI, machine learning methods that infer variant fitness from directed evolution data, identifying tighter binders missed by conventional approaches.

The paper evaluates methods for predicting protein fitness using both evolutionary and assay-labelled data, finding a simple baseline approach competitive with more complex methods.

A convolutional network trained on 16S reads from 13,838 species reaches within 2.0% of training-set Bayes-optimal species accuracy and usually beats k-mer, alignment and Kraken2 on held-out-species genera.