Automated, predictive, and interpretable inference of Caenorhabditis elegans escape dynamics

The cost of an empirical bit in biophysics has fallen dramatically, and high-precision data are now abundant. However, biological systems are notoriously complex, multiscale, and inhomogeneous, so that we often lack intuition for transforming such measurements into theoretical frameworks. Modern machine learning can be used as an aid. Here we apply our Sir Isaac platform for automatic inference of a model of the escape response behavior in a roundworm directly from time series data. The automatically constructed model is more accurate than that curated manually, is biophysically interpretable, and makes nontrivial predictions about the system.

Read this article

Bryan C. Daniels, William S. Ryu, Ilya Nemenman
Proceedings of the National Academy of Sciences Mar 2019, 201816531; DOI: 10.1073/pnas.1816531116