ADH-PPI: A Collection of Neural Refinements for Protein Protein Interaction Prediction makes use of novel most informative amino acid distribution based sub-sequence generation approach and a precisely deep neural network to outperform state-of-the-art PPI prediction performance over 2 core datasets (Human, Mouse) and 4 independent test sets (Human, Mouse, E.coli, C.Elegans). ADH-PPI web interface supports multi-dimensional analysis of protein sequences, training and optimizing the deep learning model from scratch, using pre-trained models of 2 different species to make inference on new sequences, and download interactive artifacts during the lifetime of session.