Monitoring Error Rates In Illumina Sequencing

Abstract

Guaranteeing high-quality next-generation sequencing data in a rapidly changing environment is an ongoing challenge. The introduction of the Illumina NextSeq 500 and the depreciation of specific metrics from Illumina's Sequencing Analysis Viewer (SAV; Illumina, San Diego, CA, USA) have made it more difficult to determine directly the baseline error rate of sequencing runs. To improve our ability to measure base quality, we have created an open-source tool to construct the Percent Perfect Reads (PPR) plot, previously provided by the Illumina sequencers. The PPR program is compatible with HiSeq 2000/2500, MiSeq, and NextSeq 500 instruments and provides an alternative to Illumina's quality value (Q) scores for determining run quality. Whereas Q scores are representative of run quality, they are often overestimated and are sourced from different look-up tables for each platform. The PPR's unique capabilities as a cross-instrument comparison device, as a troubleshooting tool, and as a tool for monitoring instrument performance can provide an increase in clarity over SAV metrics that is often crucial for maintaining instrument health. These capabilities are highlighted.

Keywords: bioinformatics; genomics; high-throughput DNA.

FIGURE 1.

PPR output for HiSeq 2000 and NextSeq 500. (A) PPR summary. (B) PPR plots for a 150 + 150 NextSeq 500 v1 run that clustered at a density of 1083 K/mm² and yielded 494 million reads passing filter. (C) PPR plots for a 150 + 150 NextSeq 500 v2 run that clustered at a density of 525 K/mm² and yielded 315 million reads passing filter. (D) PPR plot generated in Illumina SAV v1.8.37 for a HiSeq 2000 lane with index read shaded yellow. (E) PPR plot generated by our program for the same 100 + 8 + 8 + 100 HiSeq 2000 lane that clustered at a density of 742 K/mm2 and yielded 192 million reads passing filter.

FIGURE 2.

PPR and Illumina Q30 for v1 and v2 NextSeq 500 runs. (A) V2 chemistry data reveal a marked increase in error-free reads over v1 chemistry. Average v1 and v2 0 mismatch percentages are 59% and 71%, respectively. (B) V2 chemistry runs have higher average total percent ≥ Q30 scores and less drop-off in total percent ≥ Q30 scores with higher cluster densities. Average v1 and v2 Q scores are 80 and 87, respectively.

FIGURE 3.

PPR profiles for common sequencing failures. (A) A PPR plot for a normal run. (B) Adapter read-through failure mode. (C) Bad reagent failure mode. (D) Short repeating sequence failure mode.

FIGURE 4.

PPR profiles showing decay of Camera 1 over time.