MLCPP 2.0

MLCPP 2.0 predicts cell-penetrating peptides (CPPs) and quantifies their uptake efficiency using an interpretable stacking model that integrates multiple machine learning classifiers.

Key Features:

• Interpretable Stacking Model: integrates multiple machine learning classifiers in a stacking approach to improve prediction accuracy and interpretability.
• Comprehensive Benchmarking Dataset: uses an updated benchmarking dataset for training and validating predictive models.
• Diverse Feature Encoding Algorithms: evaluates 17 sequence-based feature encoding algorithms to capture peptide sequence characteristics.
• Multiple Machine Learning Classifiers: constructs baseline models using seven conventional machine learning classifiers.
• Systematic Model Optimization: identifies optimal feature sets and classifiers for predicting CPP presence and uptake efficiency separately.
• Cross-Validation Strategy: performs multiple 10-fold cross-validation to build 119 baseline models and merges predicted probability values into a new feature vector.

Scientific Applications:

• Discovery and design of CPPs: enables identification and design of novel cell-penetrating peptides with predicted uptake efficiency.
• Experimental hypothesis generation: provides predicted CPP status and uptake strength to inform hypothesis-driven experiments.
• Drug delivery and gene therapy research: supports selection of CPP candidates for applications in drug delivery systems and gene therapy.

Methodology:

MLCPP 2.0 updates a benchmarking dataset, evaluates 17 sequence-based feature encodings, constructs baseline models using seven machine learning classifiers, applies multiple 10-fold cross-validation to generate 119 baseline model predictions, and merges predicted probability values into a new feature vector for stacking.