Statistical prediction of the bioactivity of amoebapores peptide and its analogs

Totally 77 physicochemical properties of amino acids are collected and used as the structural descriptors of peptide sequences, and based upon it several statistical models are constructed by using stepwise multiple regression (SMR) and partial least squares (PLS) regression. In both MRL and PLS model, the obtained correlation coefficients r are all above 0.95. And then the predictive capability of these statistical models is further confirmed by using leave-one-out (LOO) cross-validation technique. The modeling results show that hydrophobicity and electronic property have a fundamental influence on the minimal growth inhibitory concentration (MIC) and minimal lethal concentration (MLC) of amoebapores peptide and its analogs.