Predicting the solubility of recombinant proteins in Escherichia coli

Abstract

We describe a statistical model that uses binomial logistic regression for predicting the solubility of heterologous proteins expressed in E. coli. The model is based on a set of proteins reported to have been expressed in E. coli in either soluble or insoluble form. The 22 parameters used in the final model based on proteins’ amino acid composition are discussed. The overall accuracy of the model developed is 94%. The way to use this model on the website http://www.ou.edu for the prediction of protein solubility is explained.