A pipeline for improved QSAR analysis of peptides: physiochemical property parameter selection via BMSF, near-neighbor sample selection via semivariogram, and weighted SVR regression and prediction

Abstract

Received: / Accepted: / Published:Abstract: In this paper, we present a pipeline to perform improved QSAR analysis of peptides. The modeling involves a double selection procedure that first performs feature selection and then conducts sample selection before the final regression analysis. Five hundred and thirty-one physicochemical property parameters of amino acids were used as descriptors to characterize the structure of peptides. These high-dimensional descriptors then go through a feature selection process given by the Binary Matrix Shuffling Filter (BMSF) to obtain a set of important low dimensional features. Each descriptor that passed the BMSF filtering also receives a weight defined through its contribution to reduce the estimation error. These selected features were served as the predictors for subsequent sample selection and modeling. Based on the weighted Euclidean distances between samples, a common range was determined with high-dimensional semivariogram and then used as a threshold to select the near-neighbor samples from the training set. For each sample to be predicted, the QSAR model was established using SVR with the weighted, selected features based on the exclusive set of near-neighbor training samples. Prediction was conducted for each test sample accordingly. The performances of this pipeline are tested with the QSAR analysis of angiotensin-converting enzyme (ACE)inhibitors and HLA-A*0201 data sets. Improved prediction accuracy was obtained in both applications. This pipeline can optimize the QSAR modeling from both the feature selection and sample selection perspectives.This leads to improved accuracy over single selection methods. We expect this pipeline to have extensive application prospect in the field of regression prediction.

Comprehensive Evaluation and Comparison of Machine Learning Methods in QSAR Modeling of Antioxidant Tripeptides

“…Therefore, the significant improvement in model performance was achieved by feature selection because plenty of irrelevant features were eliminated. 17 …”

“…comprising the peptides and properties of the entire peptides (electronegativity, sequence information, solubility, molecular weight, topological information, etc.). 17 , 18 Then, feature selection and modeling methods are combined to connect the structure information and bioactivity. 17 , 19 More than 80 amino acid descriptors (AADs) extracted from properties of amino acids by principal component analysis (PCA) were presented to characterize peptide structures and encode the peptides.…”

“…comprising the peptides and properties of the entire peptides (electronegativity, sequence information, solubility, molecular weight, topological information, etc. ). , Then, feature selection and modeling methods are combined to connect the structure information and bioactivity. , More than 80 amino acid descriptors (AADs) extracted from properties of amino acids by principal component analysis (PCA) were presented to characterize peptide structures and encode the peptides. − However, directly using these AADs usually led to undesirable model performance since most of them were not intended for the antioxidant activity modeling (e.g., T-scale for angiotensin-converting enzyme inhibitory activity). ,,,,− …”

“…Machine learning methods have been successfully applied for feature selection and model development in QSAR studies on peptide bioactivity (e.g., angiotensin-converting enzyme inhibitory activity). ,,,,− A total of 566 numerical values of amino acid including physicochemical properties and biochemical properties of amino acids and pairs of amino acids have been available in the AAIndex database . This makes it possible to use feature selection to find the important variables for bioactivity prediction compared with using AADs from PCA where the principal components were composed of various original variables.…”

Maximal Information Coefficient and Support Vector Regression Based Nonlinear Feature Selection and QSAR Modeling on Toxicity of Alcohol Compounds to Tadpoles of Rana temporaria

“…This study uses the molecule descriptor calculation software, PCLIENT, to calculate thousands of physiochemical parameters for every small molecule compound of alcohol. 15 Optimum descriptors subset is obtained by a feature selection pipeline containing three step searching strategies: (i) select statistically significant features that imply nonlinear correlation with biotoxicity of chemical compounds using MIC based univariate filter; (ii) refine feature subset by support vector regression based backward elimination (SVR-BE); 16 (iii) obtain optimal subset via a forward selection process that integrated minimal redundancy maximal relevance, MIC and SVR. A QSAR model is finally built on the training set with the reserved descriptors, and then to predict biotoxicities of Rana temporaria in the test set.…”