The first massively-parallel DNA sequencing systems to be developed used PCR-amplified DNA templates; the signals generated need to be amplified in order to enable their detection during the sequencing process. The methods used for the sequencing reactions give generally low signal-to-noise ratios (the chemistry of the sequencing reactions becomes progressively more inefficient during the reactions, which is why the read lengths are generally low). In the ABI SOLiD sequencing platforms, for example, the relatively low signal-to-noise ratio means read lengths of only 35 to 75 bp, but in the Roche/454 GS FLX+ sequencing platform a significantly higher signal-to-noise ratio allows read lengths approaching that of Sanger sequencing (see Figure 1).
Fig1. Sequence throughput and read lengths for some common massively-parallel DNA sequencing platforms. For some methods the accuracy of sequencing is low for individual sequence reads but can be compensated for by having high sequence coverage (when each region of the DNA is represented by many individual sequence reads). The maximum sequence throughput of close to 1,000,000 Mb of DNA corresponds to 1000 Gb or about 300 haploid human genomes. (Data from Reuter JA et al. [2015] Mol Cell 58:586–597; PMID 26000844.)
These methods offer a trade-off: generally high, and sometimes very high, sequencing throughput, but a quite high rate of sequencing errors in individual sequence reads. By sequencing very many PCR-amplified fragments, however, many sequence reads can be obtained for a given region of DNA. This “deep sequencing” allows sequencing errors to be identified in individual reads (see Figure2) so that a consensus sequence can be established with acceptably low error rates.
Fig2. Aligning short sequence reads to a reference sequence. Some high-throughput sequencing platforms have quite short read lengths. This highly-schematic example illustrates sequence reads from paired-ends (shown in pale green and orange colored bars); that is, sequences obtained from the ends of individual DNA fragments (see Figure 6.21 for more detail). They correspond to nucleotide sequences in the reference sequence shown at top, except for indicated base changes. (Note that the bars and connecting gray lines represent individual DNA fragments, with the gray lines representing the unsequenced central part of the fragments. DNA fragments at top left and bottom right are artificially truncated, because only part of their sequence corresponds to the reference sequence.) In this example, the bases at top are covered by at most 16 sequence reads. Vertical boxes indicate the following: (1) variants that are most likely sequencing errors; (2) regions of poor coverage—is the G in the first of these regions a sequencing error, or is the individual heterozygous with a T and a G? (3) Convincing evidence of heterozygosity (C/G). A real example would have much greater sequence coverage (read depth).
