Forecasting residue-residue contact prediction accuracy.

摘要: Motivation: Apart from meta-predictors, most of today's methods for residue-residue contact prediction are based entirely on Direct Coupling Analysis (DCA) correlated mutations in multiple sequence alignments (MSAs). These average approximately 40% correct the 100 strongest predicted contacts each protein. The end-user who works a single protein interest will not know if predictions either much more or less than 40%, which is especially problem to steer experimental research that Results: We designed regression model forecasts accuracy individual proteins with an error 7 percentage points. Contacts were two DCA (gplmDCA and PSICOV). models built parameters describe MSA, secondary structure, solvent accessibility scores target Results show our can be also applied meta-methods, was tested RaptorX. Availability implementation: All data scripts available http://comprec-lin.iiar.pwr.edu.pl/dcaQ/. Contact: malgorzata.kotulska@pwr.edu.pl. Supplementary information: at Bioinformatics online.