Bioinformatics Models And Application In Predicting Protein Solubility

Protein solubility plays a crucial role in structural and biophysical research. In manycircumstance, these studies require proteins in soluble status. Besides, upon the request ofconcerning drugs, there is also a demand in industry. Experimental approaches alwaysaccompanied by limitations, for instance, cost in both money and time. Therefore,predicting protein solubility by the use of bioinformatics models can promote the studieseffectively.This paper studied the bioinformatics models for predicting protein solubility, andselected several models to introduce and compare. Nevertheless, we analyzed and testedthese models on independent data set. In addition, we made a summary and evaluationfrom algorithms, performance, and functionality support perspectives.This paper extracted data from protein information database, selected amino acidsequence as well as some physical and chemical properties as features, and then classifiedamino acids. SVM was used to train data and build predictive model. The result of10-foldcross-validation represented capability of this model, where the accuracy reached0.759,with a sensitivity of0.761, specificity of0.757, and MCC value of0.518, indicating thatour predictive model had achieve a satisfactory performance.