creating_training_set

Fish out sequences/structures from PDB which are similar but not identical to the benchmark/test sets for training deep learning models. I am using PSI-BLAST against all PDB protein sequences to search for pdbids <= threshold sequence identity

test sets

Protein-protein Protein-DNA Protein-RNA should have both unbound and bound conformations (to account for conformational changes)

read blast output

reads blast output

Options

Short	Long	Description
-i	--input_file	Path to the input file
-o	--out_file	Path to output file
-s	--seqid	Sequence identaty cutoff
-h	--help	Prints help

Example

python3 read_blast_output.py -i ../protein_DNA_RNA_benchmarks/test_set1_results.txt -o ../protein_DNA_RNA_benchmarks/test_set.txt -s 80 

get decoys

check if structures are disimular from each other

Options

Short	Long	Description
-i	--input	Path to the input file
-t	--tm_path	Path to tmalign program
-d	--pdb_dir	Directory to PDB files (PDB files need to be preprocessed)
-s	--seqid	Sequence identaty cutoff
-r	--rmsd	RMSD cutoff
-h	--help	Prints help

Example

python3 get_decoys.py -i ../protein_DNA_RNA_benchmarks/test_set1_results -t ~/prog/TMalign -d ~/PDB/ -s 100 -r 1.0

download clean PDB

download and cleans pdbfiles

Options

Short	Long	Description
-i	--input	Path to the Blast output file
-d	--pdb_dir	Directory to PDB files (PDB files need to be preprocessed)
-s	--seqid	Sequence identaty cutoff
-h	--help	Prints help

Example

python3 download_clean_pdb.py -i ../protein_DNA_RNA_benchmarks/test_set1_results -d ~\PDB -s 100

to add

within identical/similar sequences different structures to create decoys clustering with sequence/structure similarity metrics

similarity measure TM-score or RMSD