Binding Affinity Prediction of Protein-Protein Complexes
During my final year of the integrated MSc program at the Institute of Bioinformatics and Biotechnology (IBB), Savitribai Phule Pune University, I undertook a project titled “Binding Affinity Prediction of Protein-Protein Complexes Using Machine Learning” under the guidance of Dr. Sukanta Mondal.
The research involved leveraging graph theory to model protein-protein interactions (PPIs) through amino acid networks (AANs). I explored node-weighted and edge-weighted networks to represent protein structural features. Using a dataset of 101 heterodimeric protein-protein complexes, I applied machine learning models to predict binding affinities, distinguishing high and low affinity complexes. This required integrating advanced machine learning techniques, feature engineering, and cross-validation methods to optimize predictive accuracy.
Key Contributions:
- Compared two graph-based models: NAPS (node-based) and NACEN (edge-based), identifying NACEN as the superior approach for our dataset.
- Developed a unique methodology to identify functional residues and hotspots critical for protein interactions.
- Highlighted the significance of aromatic and positively charged residues in protein binding sites.
This project enhanced my skills in computational biology, data analysis, and machine learning.