Predicts an intrinsically disordered protein linker identity for smallest RMSD of amyloid beta 42 polymerized together with hexapeptide linkers versus the experimentally-determined 2beg.pdb
(1) Run reduced_linker_set.py using slurm_reduced_linker.sh to obtain the encodings for all possible hexapeptide linkers excluding W, C, and P.
(2) Run k_means_clustering.py to obtain a set of 100 encoded linkers that are clustered by taking the closest encoding to the centroids (KMedoids too memory intensive).
(3) Can run random_sampling.py to obtain a random set of unencoded linkers.
NOTE generate_fasta_files.py can be used generally to generate fasta files relevant, but was not used in this implementation using ColabFold
(4) Run average_2beg_workup.py to average the reference model across the 10 models provided experimentally.
(5) Run AlphaFold2_Github_Submission.ipynb in a Google Colaboratory to run the prediction of structures.
(6) Run compute_rmsd.ipynb in a Google Colaboratory to calculate the RMSDs between the predicted structures and the avg_2beg.pdb.
(7) Feed rmsd values and pick acquisition function to run active_learning_loop.py and obtain new predictions to loop 5, 6, and 7.
Data analysis is performed using relevant files provided.
RESULTS The linkers selected at each active learning loops and their acquisition function values are in the active_learning_results_dict directory. The linkers and their calculated rmsd values are in the rmsd directory.