A Bayesian approach for correcting Tn5 transposition bias in ATAC-seq footprinting

ATAC-seq exploits the observation that the pattern of transposition of a hyperactive Tn5 transposase in native chromatin mirrors genome-wide chromatin accessibility. It has been suggested that transposition observed around transcription factor binding motifs can be used to assess their occupancy in the form of footprints. However, we show that the vast majority of footprints observed at transcription factor motifs in ATAC-seq data spuriously arise from the intrinsic sequence-dependent transposition site bias of Tn5 and are also observed in naked DNA. We demonstrate that the Tn5 transposition bias can be corrected using existing tools for sequence bias correction and a novel estimate of global occupancy in order to produce more reliable estimates of footprints.