AF3 Helper

Generate AlphaFold3 input JSONs and analyze AF3 predictions directly in the browser

Back to Home Browse AF3 scripts
– page views
What is this tool?

AF3 Helper brings two commonly used AlphaFold3 utilities to the browser so you can prototype workflows without a backend:

  • JSON Generator: mirror of af3_json_generator.py for building AF3 input JSON files from manual sequences or FASTA files, including validation and batch conversion.
  • Result Analyzer: browser implementation of af3_analysis_tool.py for summarizing pLDDT / PTM / iPTM / ranking scores from AF3 prediction folders.

The Python scripts remain available for automation and large-scale runs; this page offers a lightweight way to inspect and prepare small jobs online.

AF3 JSON generator
AlphaFold3 expects JSON with dialect=alphafold3, version=1, a name, and a list of modelSeeds. All sequences must contain standard amino acids (ACDEFGHIKLMNPQRSTVWY).
Shared by all sequences

Click "Add chain" to add a new sequence input, or "Remove chain" to remove the last one (max 10 chains). Each sequence requires a unique Sequence ID and amino acid sequence.

AF3 result analyzer

Upload a ZIP archive containing one or more AlphaFold3 prediction folders (each folder should contain structure files and summary_confidences.json). The analyzer will compute the average pLDDT from .cif/.pdb and extract PTM/iPTM/ranking scores, mirroring the Python CLI.

Preparing the ZIP: right-click the AF3 output directory (the one containing per-target subfolders) and compress it. Avoid huge archives; the browser must hold everything in memory.
Download scripts

Prefer running locally or in pipelines? Grab the original Python scripts used for this page.

af3_json_generator.py af3_analysis_tool.py AF3 README

Scripts located in af3/scripts/. Install dependencies via pip install -r af3/requirements.txt.

Quick start (CLI)

Generate AF3 JSON from FASTA

python af3/scripts/af3_json_generator.py --fasta sequences.fasta --output complex.json

Batch convert a FASTA directory

python af3/scripts/af3_json_generator.py --batch_input fastas/ --batch_output jsons/

Analyze AF3 predictions

python af3/scripts/af3_analysis_tool.py --input_dir ./af3_results --output_dir ./analysis
Tips
  • AF3 JSON sequence strings must contain only the 20 canonical amino acids. The validator highlights invalid characters.
  • When analyzing results, the page looks for .cif first (prioritizing filenames containing model), falling back to .pdb if needed.
  • Large archives may take noticeable time to parse. If the browser becomes unresponsive, analyze fewer samples at once or switch to the Python CLI.
Analytics dashboard

Live Google Analytics 4 metrics rendered via Looker Studio. Use the embedded controls to explore visitor behaviour for the Toolboxes suite.