2× genomes—Does depth matter?

This issue of Genome Research marks the publication by collaborators at the NIH, Agencourt Bioscience, and the Broad Institute of a genome sequence for the domestic cat Felis catus (Pontius et al. 2007). It was obtained by a whole-genome shotgun approach in which, on average, each genomic base is represented in roughly two sequence reads (“2×” redundancy), a level at which there remain many gaps in the sequence due to statistical fluctuations in read placement, biases in subclone libraries, and assembly difficulties. In all, the NHGRI is sponsoring 24 such “low redundancy” mammalian genome sequences, 17 of which have already been assembled and released (Table 1). Complete coverage of every genome would obviously be preferable; the decision to acquire multiple incomplete sequences represents a compromise balancing phylogenetic breadth (inclusion of many species) against redundancy depth for particular species. In this commentary, I discuss some of the issues involved in this compromise, and touch on the uses, characteristics, and limitations of 2× assemblies.

The tradeoff between breadth and depth has actually been a recurring theme in genome sequencing, because the choice between spreading available data-generating capacity broadly over a larger extent of targeted DNA and obtaining deeper coverage of a more limited target arises in several contexts. The appropriate balance in each case is far from obvious and depends on a number of factors, including the relative difficulty of obtaining DNA sources of various types, available data analysis tools, and the implications of sequence errors and gaps for downstream utility. Early in the Human Genome Project, the high level of redundancy required by the shotgun method was often viewed as unduly wasteful, and considerable effort was expended toward developing more efficient directed strategies. These all foundered, due to a combination of logistical complexity, the problems posed by …