By comparing the variation in chromatin accessibility across cells between highly similar kmers, we can identify critical bases associated with chromatin accessibility variation. For example, the “AGATAAG” kmer, which closely matches the GATA1 motif, is highly variable across single cells, but most kmers differing by one nucleotide share little or none of that variability (Fig. 3b, Supplementary Fig. 10). The mismatched kmer with the greatest correlated variability is “TGATAAG”, which is consistent with the weights of each nucleotide in the GATA1 motif. Similarly strong sequence specificity is seen across other variable motifs (Supplementary Fig. 10).
