Computational correction of cross-contamination due to exclusion amplification barcode spreading

Recent advances in sequencing technology have considerably increased the throughput and decreased the cost of short-read sequencing by using high-density patterned flow cells. However, high rates of cross-contamination between multiplexed libraries have been observed on data from these machines, likely due to index-switching during exclusion amplification1,2. Here, we demonstrate that a computational correction procedure based on the Sylvester equation removed 80-90% of the false positive expression signal and eliminated spurious clustering. The computational correction procedure can therefore be used to rescue aspects of affected sequence data so that researchers can take advantage of cost-effective sequencing on patterned flow cells.