State-of-the-art ab initio prediction of protein secondary structure

Porter 5 is composed of ensembles of cascaded Bidirectional Recurrent Neural Networksand Convolutional Neural Networks, incorporates new input encoding techniques and is trained ona large set of protein structures. Porter 5 achieves 84% accuracy (81% SOV) when tested on 3 classesand 73% accuracy (70% SOV) on 8 classes on a large independent set. In our tests Porter 5 is 2% moreaccurate than its previous version and outperforms or matches the most recent predictors of secondarystructure we tested.