In the article above we stress the trade-off decisions inescapably made when deciding to invest finite resources in either fully finished microbial genomes or in various levels of draft; we note that in our hands the cost differential is 3- to 4-fold (similar to3c draft base, similar to10c finished) and the speed differential is over 10-fold. We argue that draft sequencing of high quality is attainable at the quoted cost (yielding, e.g., only one or a few scaffolds per genome) whose scientific value is quite close to that of fully finished. The preceding article (C M. Fraser, J. A. Eisen, K E. Nelson, I.T Paulsen, and S. L. Salzberg, J. Bacteriol. 184:6403-6405) argues to the contrary that the difference in scientific value between draft and finished sequences is very great, if not essentially dichotomous, and that the cost difference is modest (1.3- to 1.5-fold, though with rough agreement between us as to the cost of finished data); there is clearly a large disagreement on both scores. Partly for this reason, that article also neglects the "lost-opportunity" cost to which we attach high importance. In our hands, producing finished genomes comes at the inescapable sacrifice of at least two-thirds of the number of genomes that could be produced in a high-quality draft (in a money-limited world), and we can produce the latter at least 10 times faster than the former. While we acknowledge the importance of fully finishing many key microbial genomes, the inestimable and unknown immensity of microbial diversity argues strongly to us that for the foreseeable future the bulk of microbial sequencing investment should be in high-quality draft form. But it is also our position that, as in the sequencing of larger genomes, we are just beginning to explore the cost and utility characteristics of imperfect data (e.g., expressed sequence tags, draft mammalian genomes, etc.) produced as intermediate or even final products. All aspects of this question are still open and changing, although in the next few years we should gain a very much clearer picture than we now have. But we should not, in principle, ignore the fact that our efforts are resource limited. There are vastly more sequence data of importance to science than we can conceivably afford to produce, even as drafts, over the next many years. So the only real question, we believe, is that of what mix of approaches we should employ. In this context the investment decisions are not unlike those of ordinary life: given limited resources, simply buying the best seldom yields the greatest overall return in value, and you quite regularly get much less than you pay for.
