Clinical Antibody Polyreactivity Labels and VH Sequences

86 clinical-stage antibody heavy chain variable domain sequences are labeled for binary polyreactivity. The dataset was compiled by Novo Nordisk and University of Cambridge researchers to reproduce a 2025 benchmark, based on original data published by Jain et al. in 2017.

Use Cases

• Predict binary polyreactivity labels from heavy chain variable domain (VH) sequences using sequence-based models.
• Benchmark machine learning models for antibody polyreactivity against the Sakhnini et al. 2025 results.
• Analyze relationships between VH sequence features and clinical-stage antibody polyreactivity outcomes.

Strengths

• Contains 86 antibody sequences with expert-curated polyreactivity labels.
• Data is preprocessed for direct benchmark reproduction from a 2025 publication.

Limitations

• Small sample size of 86 sequences limits model training and statistical power.
• Dataset is a curated subset of the original 137 sequences from Jain et al. 2017.

Provenance

Source

Original data from Jain et al. 2017; benchmark preprocessing by Sakhnini et al. 2025 (Novo Nordisk & University of Cambridge).

Collection Method

Biophysical measurements of FDA-approved or late-stage clinical antibodies.

Time Range

Original data published 2017; benchmark published 2025.

Freshness

Dataset was last updated on the platform in December 2025.