Figure 3
QCT tracking analysis quantifies contamination and sample mixups. (A) The experiment from Fig. 2C was analyzed for contamination using QCTs. Either 4x, 2x, 1x or 0 QCT molecules were added to 96 PCRs and sequenced on an Illumina Miseq lane using Truseq-style index pairs. Each EMI cluster for a given i5/i7 index pair was classified as contamination if (i) the read depth was below threshold (see Methods and Fig. S2), and (ii) the same EMI cluster was found to originate from another index pair. The QCT contamination fraction for a given index pair is the ratio of contaminating reads over total reads. (B) Demonstration of QCT tracking analysis to identify the source of contamination. Contaminating EMI clusters in the D707/D504 index pair (yellow box, A) are found to mostly originate from D701/D504 (left), and likewise, contaminants in D707/D504 are found to originate from D707/D504 (right). (C) QCT contamination tracking shows dual-unique indexes drastically reduce index misassignment. Both dual-unique indexes and Truseq-style index pairs were used on 120 samples prepared and sequenced in batch. Contamination source/destination is shown for all pairs of the 120 PCRs. The reactions that used dual-unique indexes (inset 1) typically had 0.006% contamination compared to 0.5% contamination using Truseq-style index pairs, particularly for D701 and D704 (insets 2a and 2b). Relatively high contamination was also observed in Truseq-style PCRs that had a D7xx index in common (example in inset 3), which is consistent with index misassignment. (D) Pairwise analysis of QCT fingerprints identifies sample mixups. Forty-eight PCRs were processed in parallel, dual unique indexed, and sequenced on a Miseq lane. The similarity of QCT fingerprints is quantified as the number of high read depth EMI clusters in common (i.e. collisions) between two reactions. The number of collisions for all pairs of PCRs is shown.
