Improving Coiled-Coil Prediction with Evolutionary Information

The coiled-coil is a widespread protein structural motif known to have a stabilization function and to be involved in key nteractions in cells and organisms. Here we show that it is possible to increase the prediction performance of an ab initio method by exploiting evolutionary information. We implement a new program (addressed here as PS-COILS) in order to take as input both single sequence and multiple sequence alignments. PS-COILS is introduced to define a baseline approach for benchmarking new coiled-coil predictors. We then design a new version of MARCOIL (a Hidden Markov Model based predictor) that can exploit evolutionary information in the form of sequence profiles. We show that the methods trained on sequence profiles perform better than the same methods only trained and tested on single sequence. Furthermore, we create a new structurally-annotated and freely-available dataset of coiled-coil structures (www.biocomp.unibo.it/ lisa/CC). The baseline method PS-COILS is available at www.plone4bio.org through subversion interface.