VDJ-REMIX: REpertoire Module Identification and eXploration

Abstract Motivation High-throughput sequencing of B and T cell repertoires provides unprecedented insights into adaptive immunity but generates high-dimensional feature sets that are challenging to interpret. Standard dimensionality reduction techniques are often suboptimal for adaptive immune receptor repertoire (AIRR) data, which exhibits multi-collinearity, heterogeneous data types, and missingness. Results Here, we present VDJ-REMIX, an R package implementing a robust, network-based framework to deconstruct complex repertoire feature matrices into biologically interpretable modules. By refactoring weighted correlation network analysis (WGCNA), VDJ-REMIX provides a tailored workflow for preprocessing, imputation, and modularization of immune repertoire data. We demonstrate its utility across diverse contexts, including autoimmunity, inflammation, and acute infection. In autoimmune patients, VDJ-REMIX identified distinct B cell signatures that stratified diseases and revealed opposing dynamic responses to B cell-depleting versus anti-proliferative therapies. In COVID-19 and non-COVID-19 sepsis patients, it distinguished disease-specific signatures from shared severe infection responses and identified modules correlated with severity. Analysis of flow-sorted B cell populations stratified by FCGR2B genotype recapitulated known tolerance defects and uncovered population-specific repertoire signatures linked to inhibitory receptor dysfunction, providing orthogonal validation of module biological coherence. Finally, applied to a single-cell multi-omics dataset of immune cells in pancreatic ductal adenocarcinoma (PDAC) combining gene expression with AIRR-seq, VDJ-REMIX recovered modules linking BCR isotype usage and clonality to cytotoxic, interferon-responsive, and regulatory immune programmes. VDJ-REMIX is a versatile tool enabling systematic exploration of immunological variation and biomarker discovery from complex immune repertoire data. Availability and implementation VDJ-REMIX is freely available at https://github.com/Bashford-Rogers-lab/vdjremix.