Table S1 - A Unified Multitask Architecture for Predicting Local Protein Properties

The table lists, for each prediction task, the per-residue per-class performance, i.e. (precision, recall, F1, total number of provided positive amino acids, true positive) averaged per cross-validation test fold, achieved via single-task training of the neural network with just the PSI-BLAST features (“Single”), multitask training just using the PSI-BLAST features (“Multi”), multitask training including the amino acid embedding (“MultiEmbed”), multitask training including the natural protein task (“All3”), and multitask training including the natural protein task with Viterbi post-processing (“All3+Vit”). (PDF)