abodybuilder3

What problem does it solve?

The skill streamlines scalable antibody structure prediction by leveraging protein language model embeddings to infer inter-residue geometries and 3D coordinates, enabling rapid analysis for research and therapeutic development.

Core Features & Use Cases

• GPU-accelerated prediction via Docker image ghcr.io/open-prophetdb/abodybuilder3:arm64-blackwell
• Batch processing for high-throughput antibody panels
• Supports paired heavy and light chain input in FASTA format for end-to-end predictions
• Use Case: Therapeutic antibody design, structural analysis, and computational immunology workflows

Quick Start

To run a single prediction, mount your input FASTA and run the Docker image with the inference script as shown. docker run --rm --gpus all -v $(pwd):/data ghcr.io/open-prophetdb/abodybuilder3:arm64-blackwell python /app/src/abodybuilder3/stages/inference.py --model language --output-dir /data/output For batch predictions, mount data directories and run: docker run --rm --gpus all -v /data:/data ghcr.io/open-prophetdb/abodybuilder3:arm64-blackwell python /app/src/abodybuilder3/stages/inference.py --model language --output-dir /data/outputs