Comparison of Three Next Generation Sequencers on Their Basecall Error Rates and Bias Patterns

We compared basecall error rates and error bias patterns for three high-throughput next generation sequencers: NovaSeq 6000, DNBSeq-T7 and Salus EVO. Their overall error rates fall in the 0.05~0.1% range: EVO (0.065%), T7 (0.068%) and NovaSeq (0.091%). All instruments show a declined accuracy for reported Phred Q values over sequencing cycles, with EVO giving the highest stability. Despite of their similar overall error rates, the three instruments show large difference in their performance on variant calls. Using a threshold of mutation frequency of 0.5%, NovaSeq called 1782 false positive variants out of 100,000 sites where as T7 and EVO called 210 and 104, respectively. Basecall error rate distributions discloses that there are significant non-random errors in the basecalls, which are the major cause for false positive variant calls. Further analysis shows that different sequencers have different error bias patterns including those sequence context-dependent and -independent. The instrument design and the polymerase used in the sequencing-by-synthesis approach determine the basecall accuracy of a sequencer.