Naïve Bayes modeling for active small molecule prediction. (a) Fitness model. (b) Photosynthesis model. (c) Motility/phototaxis model. (a-c) The model was trained on four-fifths of the active set and tested against the remaining fifth. Assessment is repeated five times and compared to random selection of active small molecules to calculate the model enrichment factor. Error bars represent standard error of five replicates.