A common strategy for identifying molecules likely to possess a desired biological activity is to search large databases of compounds for high structural similarity to a query molecule that demonstrates this activity, under the assumption that structural similarity is predictive of similar biological activity. However, efforts to systematically benchmark the diverse array of available molecular fingerprints and similarity coefficients have been limited by a lack of large-scale datasets that reflect biological similarities of compounds. To elucidate the relative performance of these alternatives, we systematically benchmarked 11 different molecular fingerprint encodings, each combined with 13 different similarity coefficients, using a large set of chemical-genetic interaction data from the yeast Saccharomyces cerevisiae as a systematic proxy for biological activity. We found that the performance of different molecular fingerprints and similarity coefficients varied substantially and that the all-shortest path fingerprints paired with the Braun-Blanquet similarity coefficient provided superior performance that was robust across several compound collections. We further proposed a machine learning pipeline based on support vector machines that offered a fivefold improvement relative to the best unsupervised approach. Our results generally suggest that using high-dimensional chemical-genetic data as a basis for refining molecular fingerprints can be a powerful approach for improving prediction of biological functions from chemical structures.
Bibliographical noteFunding Information:
We thank Benjamin VanderSluis, Wen Wang, and Maximilian Billmann for proofreading the early drafts of this article. H.S. was partially supported by the National Institutes of Health (NIH) (R01HG005084 and R01GM104975) and the National Science Foundation (NSF) (DBI 0953881). S.W.S. was supported by a NSF Graduate Research Fellowship (00039202), a NIH Biotechnology training grant (T32GM008347), and a BICB one-year fellowship. S.C.L. was supported by a RIKEN Foreign Postdoctoral Research Fellowship and a RIKEN Incentive Research Projects grant. Minoru Yoshida was supported in part by a Grant-in-Aid for Scientific Research (S) (19H05640), the Japan Society for the Promotion of Science (JSPS), and by a Grant-in-Aid for Scientific Research on Innovative Areas (18H05503) from the Ministry of Education, Culture, Sports, Science, and Technology (MEXT) of Japan. C.B. was supported by a JSPS Grant-in-Aid for Scientific Research (B) (15H04483). C.B. and Y.Y. were supported by a Grant-in-Aid for Scientific Research on Innovative Areas (17H06411) from MEXT. C.L.M. is a fellow in the Canadian Institute for Advanced Research (CIFAR) Genetic Networks Program. Computing resources and data storage services were partially provided by the Minnesota Supercomputing Institute and the UMN Office of Information Technology, respectively.