Prediction of protein aggregation tendency based on the support vector machine algorithm

Protein aggregation plays an important role in various diseases, for instance, type ۲ diabetes (T۲D),Alzheimer’s disease (AD), Parkinson’s disease (PD), prion encephalopathies, and Huntington’s disease.Moreover, it has been recognized as a field with increasing importance in the biopharmaceutical industriesbecause of its occurrence during bioprocessing steps to ensure the drug's effectiveness and decreaseassociated risks, such as increased immunogenicity. Therefore, aggregation prediction of proteins underdifferent conditions has great importance for successful biopharmaceutics’ development and theranosticapproaches. In order to predict the aggregation propensity of proteins, a machine learning method wasproposed to evaluate the aggregation propensity of hexapeptides of WALTZ-DB ۲.۰ databank, using theSupport Vector Machine (SVM) algorithm based on the sequences of segments and beta-sheet formationpropensity of residues as an intrinsic feature. To analyze the capability of the proposed method, twoparameters were considered, which are F-measure and Matthews Correlation Coefficient (MCC), owing totheir evaluative power. Finally, the applied approach was compared with the Pasta ۲.۰ server that uses similarinputs to make predictions. The mentioned parameters of the proposed method were resulted to be ۰.۸۳۰ and۰.۶۳۳ for the proposed method, and ۰.۶۸۸ and ۰.۳۸۲ for the Pasta۲ server, respectively. As a result, the newsuggested strategy superiorly evaluates the aggregation propensity, which is essential for the basic andapplied approaches.