chromVAR first computes a “raw accessibility deviation” for each motif and cell, representing the difference between the total number of fragments mapping to peaks containing the given motif and the total expected number of fragments based on the average of all input cells. This aggregation across peaks sharing a common motif yields a signal that is considerably less sparse than the signal within individual peaks, however this aggregation can also amplify technical biases between cells due to PCR amplification or variable Tn5 tagmentation conditions (Supplementary Note 1). These technical biases can lead to differences in the number of observed fragment counts between cells for a given peak set with distinct GC content or mean accessibility (Supplementary Fig. 2). To account for these technical confounders, “background” peak sets are created for each annotation, which comprise of an equal number of peaks matched for GC content and average accessibility (Supplementary Figs. 2–5; Supplementary Note 2). The raw accessibility deviations for these background peak sets are used to compute a bias corrected deviation and z-score for each annotation and cell, providing a bias-corrected
