Alphafold3 Server 🤠

Description:

This is a collection of helpful Python scripts for working with Alphafold3, FASTA/Genbank/JSON files, and related scripts! I had found it quite frustrating to work with the AF3 server (found here: https://alphafoldserver.com/), copying and pasting every FASTA sequence, so I created a set of Python3 notebooks to do work with AF3 in a high throughput manner.

NOTE

Most of these only work for protein sequences without any post-translational modifications (PTMs) including phosphorylation, glycosylation, SUMOlation, etc.

Contents:

1. FAA_TO_JSON_LOCAL_CLEAN.ipynb will convert a local .faa file to individual .json files formatted for AF3. Input: A .faa file, its file location, and an output directory Output: Formatted .json files for each protein in the .faa file in your specified directory
2. GB_DIRECTORY_TO_JSON_CLEAN.ipynb will convert a directory of .gb files to new folders in your given output file path, each containing a .json file for each individual .gb file. It will name both the folder and the file based off of the species name in the .gb file. Input: A file path to a directory of individual .gb files Output: Individual folders with one .json file each both named for the species name of an input .gb file
3. Added NC_ACCESSION_TO_JSON_NCBI_CLEAN.ipynb will download a NCBI Accession number and process it into a .json file formatted for the AF3 Server. Input: NCBI NC Accession number, your email, and an output directory. Output: .json file formatted for AF3 Server within your specified output directory.
4. Added GB_TO_JSON_UPDATE_2024_11_25.ipynb that will convert .gb files into output .json files. Works for both miRNAs and proteins (note: for miRNAs, AF will just display a small line without much structural information as they are very small). Input: Folder containing your .gb files Output: .json files in separate folders
5. Added JSON_SUM_CONFID_CLEAN.ipynb that will record, organize, clean, and store output metrics from your "fold_your_protein_summary_confidences_#.json" files in a .csv Input: Directory containing your summary confidence files (can be a parent directory of predicted AF3 models and it will walk through all of them) Output: .csv containing information on your models' 'fraction_disordered', 'has_clash', 'num_recycles', 'ptm', 'ranking_score', 'chain_iptm', 'chain_pair_iptm', 'chain_pair_pae_min', 'chain_ptm.
6. Added AF3_AVERAGE_PLDDT_JSON_CLEAN_2025_01_06.ipynb that will summarize the length of your predicted structure as well as the average plddt for your model (as the server does not readily give that). Input: Directory containing your fold_protein_name_full_data_#.json files Output: .csv with your protein name, protein length, and average plddt
7. GB_TO_FAA_WALK_CLEAN_2025_02_21.ipynb will take a parent directory of .gb files and create .faa files for each protein found in your .gb files. Useful when comparing orthologs among related species for downstream MSA. Input: Directory containing your .gb files Output: Two different files - 1 .faa file for every protein and individual isoforms and combined files of a protein as well as all of its isoforms (i.e. protein_A_alpha.faa, protein_A_beta.faa, and protein_A_combined.faa)
8. Added CIF_TO_3Di_CLEAN.ipynb that will take either .pdb or .cif files and transform them into the 3Di language used by Foldseek. Makes use of mini3Di from https://github.com/althonos/mini3di. Input: Directory of .cif/.pdb files containing only one specific protein (designed for orthologs) Output: .fas file of all 3Di-transformed .cif/.pdb files aligned in FASTA format
9. Added AF3_FILE_TYPE_SEP.ipynb to easily separate AF3 outputs by file types. Input: Directory of downloaded AF3 files. Output: A series of separate folders for "full_data", "model", "summary_confidences", and "job_request" files.

Updated 2025-03-21