Enabling interpretable machine learning for biological data with reliability scores

Abstract

Machine learning has become an important tool across biological disciplines, allowing researchers to draw conclusions from large datasets, and opening up new opportunities for interpreting complex and heterogeneous biological data. Alongside the rapid growth of machine learning, there have also been growing pains: some models that appear to perform well have later been revealed to rely on features of the data that are artifactual or biased; this feeds into the general criticism that machine learning models are designed to optimize model performance over the creation of new biological insights. A natural question thus arises: how do we develop machine learning models that are inherently interpretable or explainable? In this manuscript, we describe reliability scores, a new concept for scientific machine learning studies that assesses the ability of a classifier to produce a reliable classification for a given instance. We develop a specific implementation of a reliability score, based on our work in Sugden et al. 2018 in which we introduced SWIF(r), a generative classifier for detecting selection in genomic data. We call our implementation the SWIF(r) Reliability Score (SRS), and demonstrate the utility of the SRS when faced with common challenges in machine learning including: 1) an unknown class present in testing data that was not present in training data, 2) systemic mismatch between training and testing data, and 3) instances of testing data that are missing values for some attributes. We explore these applications of the SRS using a range of biological datasets, from agricultural data on seed morphology, to 22 quantitative traits in the UK Biobank, and population genetic simulations and 1000 Genomes Project data. With each of these examples, we demonstrate how interpretability tools for machine learning like the SRS can allow researchers to interrogate their data thoroughly, and to pair their domain-specific knowledge with powerful machine-learning frameworks. We hope that this tool, and the surrounding discussion, will aid researchers in the biological machine learning space as they seek to harness the power of machine learning without sacrificing rigor and biological understanding.