2018-05-23T21:23:47Z (GMT) by Bruno, Andrew E.

Besra is a tool for auto-classifying protein crystallization experiments. The goal is to implement a fast and accurate binary classifier for determining crystal-positive vs crystal-negative images in high-throughput protein crystallization experiments. Accuracy equal or better to that of a human would be considered a success. Current methods take upwards of ~10 hours to classify 1536 images. Speeding up the classification will allow better integration into existing expert knowledge systems/pipelines and enable robust evaluation and tuning of classification algorithms across millions of images. Extending from binary classification to n-way classification (clear, precipitate, skin, phase separation) is supported however performance has not been extensively tested. Source and binary releases are available at




GPL 3.0+