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

Abstract

Due to their multiple beneficial effects, antioxidant peptides have attracted increasing interest. Currently, the screening and identification of bioactive peptides, including antioxidative peptides based on wet-chemistry methods are time-consuming and highly rely on many advanced instruments and trained personnel. Quantitative structure–activity relationship (QSAR) analysis as an in silico method can be more efficient and cost-effective. However, model performance of QSAR studies on antioxidant peptides was still poor due to limited attempts in model development approaches. The objective of this study was to compare popular machine learning methods for antioxidant activity modeling and screening of tripeptides and identify the critical amino acid features that determine the antioxidant activity. 533 numerical indices of amino acids were adopted to characterize 130 tripeptides with known antioxidant activity from the published literature, and then 7 feature selection strategies plus pairwise correlation were used to screen the most important indices for antioxidant activity and model building. 14 machine learning methods were used to build models based on the feature selection strategies, respectively. Among the 98 models, non-linear regression methods tended to perform better, and the best model with an R 2 Test of 0.847 and RMSE Test of 0.627 for tripeptide antioxidants was obtained by combining random forest for feature selection and tree-based extreme gradient boost regression for model development. Based on the predicted antioxidant values of 7870 unknown tripeptides, potentially high antioxidant activity tripeptides all have a tyrosine, tryptophan, or cysteine residue at the C-terminal position. Furthermore, the predicted antioxidant activity of six synthesized tripeptides was confirmed through experimental determination, and for the first time, the cysteine or tyrosine residue at the C-terminal was found to be critical to the antioxidant activity based on both QSAR models and experimental observations.

Computer-Aided Approaches for Screening Antioxidative Dipeptides and Application to Sorghum Proteins

“…Feature importance of these variables in ABTS radical scavenging capacity model development and ORAC model development showed that N-terminus residues played a more important role in antioxidant activity. In the study of Chen et al, the variable importance in projection (VIP) values revealed that the C-terminus had greater contribution to the ABTS radical scavenging capacity of tripeptides, and such an observation was also confirmed in the study of Du et al It is interesting that antioxidant dipeptides showed a different pattern where the importance of the N-terminus was greater than that of the C-terminus. This is consistent with the VIP values from the partial least-squares regression (PLSR) models in Zheng et al, although they instead further emphasized the importance of tyrosine and tryptophan at the N-terminus .…”

“…Technically, R 2 CV could not describe the model’s performance in an unknown data set because the data used to obtain R 2 CV were used in model building . In this study, a much larger data set was collected from the latest publications, and feature selection with nonlinear regression modeling approaches was employed to link the data set and antioxidant activity, which all contributed to the better performance in test data sets. ,,,, …”

“…Feature selection was used to further screen the important features particularly for antioxidant activity prediction: therefore, significantly simplifying the model complexity. In addition, it can also shed light on the most important features contributing to antioxidant activity . These samples with 2 n variables and corresponding Y -vector in each activity data set were modeled by extreme gradient boosting (XGboost) regression.…”

“…Developing natural antioxidants has gained expanding interest because of the increasing health concerns for synthetic antioxidants and strict regulation on their usage. − Antioxidant peptides derived from food proteins have captured worldwide attention due to their advantages such as naturally sourced, better sustainability, and no or low toxic effects. ,, Generally, antioxidant peptides are released from parent proteins by enzymatic hydrolysis, fermentation, chemical hydrolysis, germination, and/or ripening and then screened by laborious chemistry methods (e.g., fractionation, isolation, purification, identification, and characterization). , However, the conventional wet-chemistry methods are time-consuming and rely highly on many advanced instruments and equipment. ,, The chemical synthesis of peptides is an alternative approach for producing and screening potentially highly active peptides. − Nonetheless, it is practically impossible to synthesize all the peptides for antioxidative peptide screening, considering the cost of synthesis and a large number of theoretically possible peptides: i.e., 400 dipeptides, 8000 tripeptides, 160000 tetrapeptides, etc. ,, …”

Convolutional Neural Network Model Based on 2D Fingerprint for Bioactivity Prediction

“…Extraction of the structural activity relationship (SAR) [ 1 , 2 ] information from chemical datasets relies on the pairwise structural comparison of all toxicophore features and small molecules, which highlights the degree of the structural relationship between the compounds [ 3 , 4 , 5 , 6 ]. The Quantitative Structure–Activity Relationship (QSAR) can correlate the compound’s chemical and structural features with its physicochemical or biological properties.…”