This integrated homolog indentifying framework uses sequence-based tools BLAST, and MMseqs2; structure-based tool Foldseek, and protein language model based tool PROST. All the homologs after all-vs-all exhaustive search for each tool and reciprocal hits filteration is given in integrated_paralogs_methods.tsv.

All the human protein structures were downloaded from AFDB --> https://alphafold.ebi.ac.uk/ Follow the steps used in integrated_paralogs_framework.ipynb to get the putative human paralogs for provided list of proteins of interests as reference set.

Follow the steps used in functional_predictions.ipynb to predict active site residues for structurally similar proteins based on the active site residue informations of previously annotated reference protein.

reproduce the virtual environment as conda create -f human_paralogs_environment.yml

Follow the FoldMason github page to install the tool @ https://github.com/steineggerlab/foldmason