Venter and his crew—scientific and nautical—took samples of about 200 liters of surface sea water from four places in the Sargasso Sea, a part of the Atlantic Ocean off Bermuda that they describe as "one of the best-studied and most well-characterized regions of the global ocean." They filtered the water and discarded items above 3 micrometers and below 0.1 micrometers in diameter. This left essentially the bacteria, which they treated as one big bucket of DNA. They then applied the shotgun sequencing technique made famous by Venter, which breaks the DNA into random chunks, inserts the chunks into bacteria to form a living library that can store and bulk up the DNA chunks, and sequences every chunk. Brute computing power fits the chunks together into larger and larger assemblies of overlapping sequences. Further computing power derives meaning by comparing the assembled sequence with existing sequences stored in databases.

The numbers are mind-boggling. The samples yielded 1.045 billion base-pairs of non-redundant sequence. In there were some 1.2 million previously unknown genes, roughly 10 times more genes than were represented in the SwissProt database at the time. Those genes came, in aggregate, from about 1,800 species. How did the researchers arrive at this number, given that they made absolutely no effort to identify the species by culturing them?

One of the crucial factors with the shotgun approach to sequencing a single species is the depth of coverage—that is, how many different times a stretch of DNA is sequenced from independent chunks. The more times a sequence is independently read, the more reliable the assembly and the more rapidly it can be constructed. In the Sargasso Sea samples, the depth of coverage was a proxy for the number of different species. A very common species would be over-represented, so its sequence would be present in greater depth than a less common species. Rare species might not be present at all. That proved the key to estimating the number of species.

From the actual depth of coverage, which could be calculated for every position in the assemblies, and the average size of the assembly, which would be larger for species with "deeper" coverage, they could estimate the number of species that contribute to the overall genome of the whole Sargasso Sea. The most conservative of these estimates suggests there may be 1,800 species, but this is very much a lower limit. If slightly lower abundances are being captured in the sequences, then the samples may contain 45,000 species or more.

Detailed scrutiny of the data reveals more of interest. For example, half of the larger assemblies, those that included 50 or more fragments, were from species that were abundant at some sampling sites but not others. This patchiness is well-documented for larger marine plankton, but not for bacteria. It might reflect an underlying patchiness in an important resource, such as the fecal pellets of zooplankton called copepods. Bacteria colonize this "marine snow," as it is known, and the filters would easily have separated these bacteria from their substrate. The sequence also reveals aspects of the evolution of marine bacteria and much more besides.