# AbBiBench: Antibody Binding Affinity Benchmark

## Facts

- Provider: AbBiBench authors
- License: CC-BY-4.0 (https://zenodo.org/api/records/16557372)
- Size: 35 files
- Format: CSV (in sampled files)
- Stored at: Hugging Face, https://huggingface.co/datasets/AbBibench/Antibody_Binding_Benchmark_Dataset
- Access: public, readable without an account
- Category: Drug Discovery
- Source: https://doi.org/10.5281/zenodo.16557372
- Page: https://scigantic.com/archive/1bad94bd-f336-5f5d-b526-088bfc1fcc5c
- Keywords: Drug Discovery, protein structure, protein language model, benchmark, CSV, Hugging Face

## About

AbBiBench (Antibody Binding Benchmarking) is a benchmarking framework for optimizing antibody binding affinity. The dataset collects sequences of antibody mutants with experimentally measured binding affinity values, together with structures of the wild-type antigen-antibody complexes, so that sequence-based and structure-based models (protein language models, inverse folding models, structure-conditioned scorers) can be evaluated on how well their scores track measured affinity. Measured 2026-11-14: binding_affinity/ holds 17 CSV files of mutant heavy or light chain sequences with affinity labels for individual antibody-antigen systems (for example the anti-VEGF 5A12 set and several AAYL antibodies), about 59 MB, and complex_structure/ holds 13 PDB files of the corresponding complexes; metadata.json and a Croissant croissant.json describe the files. Read the CSVs with pandas and the PDB files with Biopython, Gemmi or PyMOL. The authors deposited FoldX buildModel structures of the mutant complexes separately on Zenodo (record 16557372, about 8.2 GB), which links back to this repository. Useful for zero-shot mutation effect scoring, antibody affinity maturation in silico and therapeutic antibody design. The Zenodo record from the same authors is licensed CC BY 4.0; cite AbBiBench.

## Load the data

```python
import pandas as pd
df = pd.read_csv("hf://datasets/AbBibench/Antibody_Binding_Benchmark_Dataset/binding_affinity/1mhp_LC_benchmarking_data.csv")
df.head()
```

## Data structure

File types in sampled files: CSV

Example file: `binding_affinity/1mhp_LC_benchmarking_data.csv`

## Open it

- In a hosted notebook with the data mounted: https://scigantic.com/archive/1bad94bd-f336-5f5d-b526-088bfc1fcc5c
- Full record as JSON (schema card included): https://api.scigantic.com/api/archives/1bad94bd-f336-5f5d-b526-088bfc1fcc5c
- Over MCP: `get_schema_card` with id `1bad94bd-f336-5f5d-b526-088bfc1fcc5c` at https://api.scigantic.com/mcp

## Related topics

- [CSV datasets](https://scigantic.com/topics/csv)
- [Benchmark datasets](https://scigantic.com/topics/benchmark)
- [Drug Discovery datasets](https://scigantic.com/topics/drug-discovery)
- [Protein structure datasets](https://scigantic.com/topics/protein-structure)
- [Protein language model datasets](https://scigantic.com/topics/protein-language-model)
