A reliable and robust classification pipeline for protein crystallization imaging

Description

The hullabaloo surrounding the recent successes of deep learning in image processing often falls short of the results obtained when applied to real datasets—that is, without significant development on the part of the machine learning practitioner. Functional accuracies and reliable predictions are possible, but not without a mindful approach to pipeline development.

Building machine learning pipelines that combine the various technologies available to today’s data scientist in a robust and repeatable manner is the core requirement when deploying automated image processing software. But with so many options, how can we ensure our data pipeline is accurate and that our deep tech is reliable?

Christopher Watkins explores the development of a protein crystal image classification pipeline that has been autonomously deployed at CSIRO’s Collaborative Crystallisation Centre. Protein crystallization is at the heart of understanding a protein’s structure and function. As such, it is a core piece in the development of new drugs and vaccines as well as understanding the inner workings of greater biological systems. Christopher covers the trade-offs made when choosing machine learning models and the techniques used to discriminate between them and outlines an approach to reduce the effects of overfitting and account for the effects of temporal drift the data may present, in an online manner.