AIPpred: Sequence-Based Prediction of Anti-inflammatory Peptides Using Random Forest.

Area of Science:

• Computational biology
• Peptide therapeutics
• Bioinformatics

Background:

• Therapeutic peptides are crucial for inflammatory and autoimmune diseases.
• Wet-lab identification of anti-inflammatory peptides (AIPs) is resource-intensive.
• Novel computational approaches are needed for efficient AIP candidate screening.

Purpose of the Study:

• To develop a computational method for predicting anti-inflammatory peptides (AIPs).
• To identify optimal features for AIP prediction using machine learning.

Main Methods:

• Proposed a random forest (RF)-based prediction model, AIPpred.
• Systematically analyzed various sequence compositions (amino acid, dipeptide, physicochemical properties).
• Employed a feature selection protocol on dipeptide composition (DPC) for model optimization.

Main Results:

• AIPpred achieved an AUC of 0.801 in 5-fold cross-validation, outperforming a full DPC RF model.
• The optimized feature set improved prediction efficiency.
• AIPpred demonstrated superior performance on an independent dataset (AUC=0.814) compared to existing methods.

Conclusions:

• AIPpred is an effective tool for predicting anti-inflammatory peptides.
• The method can accelerate the development of AIP therapeutics.
• AIPpred supports biomedical research by enabling rapid candidate identification.