The PCR methods described are designed to reduce the effect of size- and GC-bias from the library construction process. These methods are particularly useful for ATAC-seq because of the highly diverse fragment sizes. We find that samples containing an excess of large fragments (>1 kb) are relatively hard to quantify and result in reduced clustering efficiencies when sequencing. If a library is enriched for long fragments, library fragmentation can be optimized using more or fewer cells per reaction. Alternatively, size selection prior to sequencing might be carried out to eliminate these potentially confounding long fragments. Although it is common to size-select a narrow interval for sequencing, we recommend excising a large fragment size window of 100–1000 bps to maintain high library complexity, and enable the richness of the inferences that can be extracted from the full fragment size distribution. The described PCR method has the additional benefit of serving as an estimate for library complexity. We find that if >6 additional cycles are needed (>11 total cycles) library complexity becomes a concern. Library complexity can be improved by optimizing the input cell number or by making libraries of technical replicates.
