JuxtaBAM: a species and SNP-independent method to validate the integrity of joint epigenomic and transcriptomic profiling datasets

Integrative analysis of regulatory and transcriptional datasets has become an effective strategy to identify disrupted molecular pathways contributing to common human disease. As these epigenomic datasets grow in size, the statistical support for causal inference increases, but the possibility of mislabeled samples becomes significant and can confound downstream analysis. Previously published tools for detecting sample swaps generally rely on predetermined lists of population-based single nucleotide polymorphisms (SNPs) or linkage disequilibrium maps, limiting their applicability to model organisms and to the specific genomic regions interrogated by each assay. We present a method using genetic variation within diverse cell line and tissue sample sets to identify sample inconsistencies and correct mislabeled samples. Our method, JuxtaBAM, identifies single-nucleotide variants (SNVs) directly from the high-signal regions that overlap in the input BAM files, thereby focusing genotyping on regions with sufficient read coverage in both assays. JuxtaBAM computes pairwise SNV mismatch rates across all samples, models the mismatched genotype counts with a beta-binomial distribution to test for sample mislabeling and duplication, and reports flagged pairs indicating possible sample mismatch or duplication. We demonstrate that this method can be used to validate the integrity of joint epigenomic and transcriptomic profiling datasets by applying it to data from our own previous publication on a panel of 25 gastric cancer cell lines with paired ATAC-seq and RNA-seq, in which JuxtaBAM identified a mislabeled sample that was a duplicate. We further validate the method on ENCODE mouse DNase-seq and RNA-seq data from two cell lines derived from different mouse strains, including multiple experiments generated by independent laboratories, confirming species-independent operation across laboratory sources and assay types. JuxtaBAM provides a species-independent approach for validating sample identity in multi-omic datasets, independent of predefined population SNP panels. Its reliance on variants discovered directly from the data makes it readily applicable to any organism with a reference genome, and its statistical framework provides quantitative support for identifying mislabeled samples.