Actually I'm going to evaluate the protein_protein interactions and find the hot spots using machine learning technique. Any decent references and/or suggestions for me as a beginner?
Haddock does a great job by mechanistic methods. But if you really want machine learning instead, see AlphaFold. The approach it applies to predicting the fold of a single protein could also be applied to docking of two proteins.
Machine learning techniques find the ``hotspots'''-extrema-of a cost function by AVOIDING the calculation of the interaction locally. For proteins this means focusing on their shapes and finding a way of classifying these.