Using attribution to decode binding mechanism in neural network models for chemistry
Proceedings of the National Academy of Sciences of the United States of America
National Academy of Sciences
MetadataShow full item record
Colwell, L. (2019). Using attribution to decode binding mechanism in neural network models for chemistry. Proceedings of the National Academy of Sciences of the United States of America, (201820657)https://doi.org/10.1073/pnas.1820657116
Deep neural networks have achieved state-of-the-art accuracy at classifying molecules with respect to whether they bind to specific protein targets. A key breakthrough would occur if these models could reveal the fragment pharmacophores that are causally involved in binding. Extracting chemical details of binding from the networks could enable scientific discoveries about the mechanisms of drug actions. However, doing so requires shining light into the black box that is the trained neural network model, a task that has proved difficult across many domains. Here we show how the binding mechanism learned by deep neural network models can be interrogated, using a recently described attribution method. We first work with carefully constructed synthetic datasets, in which the molecular features responsible for “binding” are fully known. We find that networks that achieve perfect accuracy on held-out test datasets still learn spurious correlations, and we are able to exploit this nonrobustness to construct adversarial examples that fool the model. This makes these models unreliable for accurately revealing information about the mechanisms of protein–ligand binding. In light of our findings, we prescribe a test that checks whether a hypothesized mechanism can be learned. If the test fails, it indicates that the model must be simplified or regularized and/or that the training dataset requires augmentation.
M.P.B. gratefully acknowledges support from the National Science Foundation through NSF-DMS1715477, as well as support from the Simons Foundation. L.J.C. gratefully acknowledges a Next Generation fellowship, a Marie Curie Career Integration Grant (Evo-Couplings, 631609), and support from the Simons Foundation. F.M. performed work during an internship at Google.
Simons Foundation (598399)
European Commission (631609)
External DOI: https://doi.org/10.1073/pnas.1820657116
This record's URL: https://www.repository.cam.ac.uk/handle/1810/293183
All rights reserved