Thethree-dimensional(3D)structurepredictionofproteinsisanimportanttaskinbioinformatics.Findingenergyfunctionsthatcanbetterrepresentresidue-residueandresidue-solventinteractionsisacrucialwaytoimprovethepredictionaccuracy.Thewidelyusedcontactenergyfunctionsmostlyonlyconsiderthecontactfrequencybetweendifferenttypesofresidues;however,wefindthatthecontactfrequencyalsorelatestotheresiduehydrophobicenvironment.Accordingly,wepresentanimprovedcontactenergyfunctiontointegratethetwofactors,whichcanreflecttheinfluenceofhydrophobicinteractiononthestabilizationofprotein3Dstructuremoreeffectively.Furthermore,afoldrecognition(threading)approachbasedonthisenergyfunctionisdeveloped.Thetestingresultsobtainedwith20randomlyselectedproteinsdemonstratethat,comparedwithcommoncontactenergyfunctions,theproposedenergyfunctioncanimprovetheaccuracyofthefoldtemplatepredictionfrom20%to50%,andcanalsoimprovetheaccuracyofthesequence-templatealignmentfrom35%to65%.
