Enhancing the Prediction of IL-4 Inducing Peptides Using Stacking Ensemble Models

Abstract

Interleukin-4 (IL-4), known for its powerful anti-inflammatory properties, plays a crucial role in immunological control. Although IL-4 is widely recognized for its ability to reduce inflammation, it can exhibit unexpected pro-inflammatory characteristics in some circumstances. The fact that IL-4-induced peptides need to be identified is of utmost relevance due to their dual nature. Our study presents a solution to the problems related to manual identification. We develop Stacking, an ensemble learning approach based on stacking, specifically built for accurately and efficiently identifying IL-4- inducing peptides. In order to improve the precision of IL-4 peptide identification, we investigate 7 Feature Extraction approaches that are based on Amino-Acid Composition (AAC), Amino-Acid Pair Composition (APAAC), Composition of k-spaced Amino Acid Pairs (CKSAAP), Composition of Tripeptide (CTDC), Dipeptide Composition (DPC), Moran autocorrelation descriptor (Moran), Pseudo Amino Acid Composition (PAAC). The Stacking model utilizes five optimized base learners (LGBM, RF, SVM, Decision Tree, KNN) and a Logistic Regression meta-learner to obtain a remarkable recognition rate of 89.97%, an MCC (Matthews Correlation Coefficient) of 0.7994, and a specificity of 0.8976.