Real-time exploration of the billion-scale human repertoire for precursor frequency analysis and rational antibody design

Abstract

The success of germline-targeting vaccines and rational antibody design depends on quantifying the population-level frequencies of specific antibody features, such as CDR lengths, sequence motifs, germline usage or preferred chain pairings, across the natural human repertoire. The maturation of high-throughput sequencing has generated large database repertoires containing billions of sequences, allowing researchers to assess naïve B-cell precursor frequencies and evaluate biologically plausible chain combinations. However, leveraging this scale of data remains a significant bottleneck, as current databases trap scientists between computationally demanding bulk downloads, restrictive, single-purpose web interfaces or outdated datasets. To address this gap, we introduce ABHunter, an automatically updated, open-access framework designed to explore and filter the healthy-human subset of the Observed Antibody Space (OAS) database, which currently stands at ~1.86-billion-sequence as of January 2026. Accessible via any web browser, the platform executes searches with median times ranging from 3.7 seconds for simple gene usage profiles to 20.8 seconds for complex paratope motifs, with full sequence retrieval completed in 118 seconds for 1 million hits. The interface allows users to quantify antibody feature frequencies to assess the natural abundance of targeted B-cells across various isotypes (e.g., utilizing IgM for precursor B-cell analysis), as well as map native repertoire contexts to scaffold isolated features. Additionally, it predicts biologically plausible inter-chain pairings for therapeutic design by using a statistical co-occurrence model that extrapolates pairing probabilities from the limited natively paired subset comprising 0.03% of the data across six donors. Furthermore, every search generates a comprehensive statistical landscape, detailing underlying complementarity-determining regions length distributions, amino acid compositions, and gene coupling patterns. For custom analysis and integration into bioinformatic pipelines, the optimized data structure reduces the storage footprint to 106 GB, allowing the entire tool to run locally on a standard laptop. ABHunter eliminates the compromise between accessibility and flexibility, transforming large-scale sequencing data into a universally accessible resource for next-generation therapeutic discovery. The platform is freely available at abhunter.org .