Analysis of the quality and utility of random shotgun sequencing at low redundancies

Abstract

The currently favored approach for sequencing the human genome involves selecting representative large-insert clones (100-200 kb), randomly shearing this DNA to construct shotgun libraries, and then sequencing many different isolates from the library. This method, entitled directed random shotgun sequencing, requires highly redundant sequencing to obtain a complete and accurate finished consensus sequence. Recently it has been suggested that a rapidly generated lower redundancy sequence might be of use to the scientific community. Low-redundancy sequencing has been examined previously using simulated data sets. Here we utilize trace data from a number of projects submitted to GenBank to perform reconstruction experiments that mimic low-redundancy sequencing. These low-redundancy sequences have been examined for the completeness and quality of the consensus product, information content, and usefulness for interspecies comparisons. The data presented here suggest three different sequencing strategies, each with different utilities. (1) Nearly complete sequence data can be obtained by sequencing a random shotgun library at sixfold redundancy. This may therefore represent a good point to switch from a random to directed approach. (2) Sequencing can be performed with as little as twofold redundancy to find most of the information about exons, EST hits, and putative exon similarity matches. (3) To obtain contiguity of coding regions, sequencing at three- to fourfold redundancy would be appropriate. From these results, we suggest that a useful intermediate product for genome sequencing might be obtained by three- to fourfold redundancy. Such a product would allow a large amount of biologically useful data to be extracted while postponing the majority of work involved in producing a high quality consensus sequence.

Figure 1

Contig formation at lower redundancy of sequencing. The number of contigs that were larger than 2 kb was calculated for each low redundancy simulation. The fold redundancy of each clone was calculated based on the number of bases that had a Phred value >20. The projects that were examined are listed at right.

Figure 2

Areas containing shallow depth at lower-redundancy sequencing. The lower redundancy-simulated sequences were tested for the occurrence of areas of shallow depth of coverage (see text). The projects examined are listed at right.

Figure 3

Assessment of consensus base quality at lower redundancies of sequencing. The Phrap value generated for each consensus base was examined, and the total number of bases that had a value above 40 were counted. The number of bases containing values >40 is represented as a percentage of the total number of bases in the project. Low values at the termini of contigs were excluded from the totals. A bar (dashes) at the 90% level is shown for reference purposes.

Figure 4

Alignment of contigs generated from low-redundancy sequencing to the known sequence. The lower-redundancy consensus sequences were compared to the completed sequence that was submitted to GenBank. The distance from the start of the project is indicated at the bottom, and the percent identity is indicated at left. Exons (numbered solid boxes) and repeats [SINEs other than mammalian interspersed repeats (MIRs) are light gray triangles pointing toward the A-rich 3′ end; LINE1s are open triangles; MIR and LINE2 elements are solid triangles] are indicated at the top.

Figure 5

Contiguity of genes on project J19 at different levels of redundancy. The number of contigs at each coverage that described a gene was counted. (Top) The genes that contain 10 or more exons and are therefore termed complex; (bottom) the contiguity of genes that contain <10 exons are therefore termed simple. The names of the genes are indicated at right; the number of exons in each gene follows the name.

Figure 6

Contiguity of genes on project P7 at different levels of redundancy. The number of contigs at each coverage that described a gene was counted. (Top) The genes that contain 10 or more exons; (bottom) the genes that contain <10 exons. The names of the genes are indicated at right; the number of exons in each gene follows the name.

Figure 7

Interspecies comparisons at different levels of redundancy. The consensus sequences generated from low-redundancy sequencing simulations were aligned to the completed sequence. (A) Percent identity between the simulated consensus and the finished sequence is shown for a series of coverages. The fold redundancy is indicated at left, the percent identity is indicated at right, and the distance from the beginning of the project is indicated on the bottom. (Top) The exons and repeats are indicated (see legend to Fig. 4 for description). (B) The number of highly homologous regions that exists between the human and mouse projects was counted. The percent of these regions that were sequenced is indicated (Sequenced), as is the percent of these regions that were identified by a homology search (Identified).