Protein binding affinity with graph features

The problem

Protein interactions depend on structural relationships that are difficult to capture with simple sequence summaries. Graph-based representations provide a way to examine those relationships.

My contribution

I investigated node- and edge-weighted amino acid networks, compared NAPS and NACEN representations, and explored machine learning models for affinity prediction.

• Engineered graph-based representations of protein complexes.
• Examined residues and interaction hotspots.
• Used cross-validation to compare candidate modeling approaches.

What this demonstrates

Domain-informed feature engineering, comparative method evaluation, and translating a biological question into a computational analysis.