HemoPImod

HemoPImod predicts the hemolytic, hemotoxic, or red blood cell (RBC) lysing potential of chemically modified peptides to support safety assessment in peptide-based therapeutics and diagnostics.

Key Features:

• Comprehensive Dataset: Uses a balanced dataset comprising 583 modified hemolytic peptides and 583 non-hemolytic peptides for model development.
• Machine Learning Approach: Employs Random Forest classifiers trained on descriptors including 2D and 3D structural features, fingerprints, atom compositions, and diatom compositions.
• Predictive Performance: Achieves 78.33% accuracy and AUC 0.86 on the main dataset, and 78.29% accuracy and AUC 0.85 on the validation dataset.

Scientific Applications:

• Drug discovery for peptide-based therapeutics: Identifies peptide modifications with potential hemolytic risk early in lead optimization.
• Peptide design and optimization: Guides selection of chemical modifications to minimize RBC lysis and hemotoxicity.
• Safety assessment in diagnostics and treatments: Evaluates hemolytic potential of modified peptides used in medical diagnostics and therapeutic applications.

Methodology:

The dataset is divided into training and validation subsets; features extracted include 2D and 3D structural descriptors, fingerprints, atom compositions, and diatom compositions; Random Forest classifiers are trained and evaluated on these features.