A new machine learning tool that participated in the CAGI 6 experiment to predict if protein single residue mutations are deleterious or benign. This was achieved in a single sequence-based approach, without multiple sequence alignments or structural information. To substitute for alignment information we used global information. The human data for training was taken from ClinVar. Testing on post-training data from ClinVar gives high area under the receiver operating curve (AUC) and the Matthews correlation coefficient (MCC). However, for genes for which the training data is either sparse or unbalanced, the prediction accuracy is poor. The resulting prediction server is available online at http://www.mamiris.com/Shoni.cagi6.

MAMI/RIS Research and Information Systems, LLC (All Rights Reserved).