The Next Generation Becomes the Now Generation

Abstract

In recent years, several so-called next-generation DNA sequencing platforms have begun to challenge the well-established Sanger sequencing method. In two important ways—cost and speed—these next-gen technologies provide improvements over Sanger sequencing. Several technical drawbacks (short read length, lack of paired end reads, and quality problems, particularly with homonucleotide stretches [1]), however, render assembly difficult and limit the use of post-Sanger sequencing. These obstacles limited the effective use of next-generation sequencing to the sequencing of prokaryotes [2], the resequencing of individuals [3], and transcriptomics studies, recently termed RNA-Seq [4] and effectively precluded de novo eukaryotic sequencing. Realizing the shortcomings of next-generation technology, manufacturers have continued to improve the read length and have recently implemented paired end methods. Capitalizing on these improvements, the publication by Nowrousian et al. describes the team's success in completely bypassing Sanger sequencing to produce a de novo assembly (to draft quality) of a complete genome, that of the filamentous fungus Sordaria macrospora [5], using Solexa sequencing-by-synthesis and 454 pyrosequencing. The technical merits of this publication make it an excellent starting point for future genome sequencing using post-Sanger platforms. The assembly phase has been a particular sticking point for de novo genome sequencing in eukaryotes, as the complexity of the genomes makes it difficult to correctly place short reads. By sequencing to high depth (nearly 100 times the length of the genome), the authors were able to pull the assembly together in large pieces (contigs) and obtain a reasonable N50 = 117 kb (defined as the smallest length of the longest contigs that cover 50% of the genome). The authors also experimented with different levels of coverage and different combinations of reads to produce assemblies of various qualities. They determined that the depth to which S. macrospora was sequenced may not be necessary, and that closing gaps with 454 reads resulted in a large improvement. Interestingly, this is similar to the blend of long- and short-insert libraries that were used for the whole genome shotgun version of the human genome project [6]. By leveraging the short inexpensive Solexa reads for the bulk of the genome, the longer 454 reads can add valuable contig order and orienting information and vastly improve quality while dramatically reducing the associated cost. Nowrousian et al. [5] have provided the assembly statistics for various depths and platforms, paving the way for future studies using high throughput sequencing. The researchers also showed that post-Sanger sequencing technologies can be used to reliably assemble difficult areas of the genome. One region of the genome, that which controls nonself recognition, could have been a particularly troublesome stumbling block. Anastomosis is a process by which hyphae, the thread-like projections of filamentous fungi, fuse and bring genetically distinct nuclei into contact. Fungi from the same species with different het (heterokaryon incompatibilty) loci will fuse, but the resulting heterokaryotic cells are subject to either severely restricted growth or cell death. This process has benefits that the authors describe briefly. Although incompatibility has never been observed in S. macrospora, the investigators report that the genome contains apparent heterokaryon incompatibility genes, with the twist that the region is inverted and contains duplications of key genes near the ends of the inversion. Such a duplication might be difficult to resolve with short Solexa data and even the longer 454 reads. However, the authors used polymerase chain reaction (PCR) to amplify across the boundaries of the inverted and duplicated region, and end-sequenced the PCR products to confirm the genome structure predicted by the genome assembler Velvet [7]. Given this demonstrated success in resolving a difficult region containing duplicate genes, researchers and physicians can consider the previously unfeasible next-gen sequencing technologies when deciding whether to sequence an entire genome. The quality of sequence produced, and ability to compare the Sanger and post-Sanger sequence scores, were additional sticking points to relying completely on the lower cost next-gen technologies. On this front, Nowrousian's team gave us a glimpse of the error rate and how it compares to that of Sanger sequencing by choosing several possible frame shifts in predicted coding regions for resequencing. The outcome of this investigation, although based on a small (21 kb total) sample, shows that the next-gen technologies can achieve error rates similar to those of Sanger sequencing. This leaves no obvious reason to use any Sanger sequencing for future whole genome sequencing projects.