Genome sequencing: mind the gap

The year 2000 is poised to be declared the ‘Year of the Genome’. The sequences of human chromosomes 21 and 22 have been published in the past six months, and two further landmark announcements are expected within days — indeed, may even have been made by the time you read this. Scientists of the Human Genome Project are on the verge of announcing completion of a “rough draft” of the human genome. And the world should soon hear another stunning proclamation from Celera Genomics, which recently completed a genome draft and is now assembling its data into near-final form. But amid all the bluster about who will deliver the sequence first, there’s been little attention paid to the quality and completeness of the data we’ll have after the sensational announcements are made and the hoopla’s died down. What, for instance, is meant by a ‘rough draft’? The rough draft from the Human Genome Project — an international consortium of labs funded primarily by the US government and the UK’s Wellcome Trust — will sequence 90% of the roughly 3.2 billion bases of the human genome. Celera says its soon-to-be-finished product (for which it is also using the public-domain data) will be 99% complete. The criteria for completion of these stages are arbitrary. Finishing the job will be the hardest part, a matter of diminishing returns, and 1–3 years may be needed to tackle those remaining percentage points by closing the thousands of gaps in the sequence. Which raises the question: Are the not-quite-finished data we’re about to see reliable, and useful for biomedical research? Or are the numerous gaps in the sequence a tiny problem with huge effects? In answering these questions, it’s important to know why the gaps exist at all. There are two types of gap: one is a statistical artifact, the other is biologically meaningful. Statistical gaps are a byproduct of the ‘shotgunning’ sequencing methods used by both the public and private efforts. DNA is chopped up into small pieces randomly, and the pieces are sequenced enough times to cover any particular point on the genome several times over, on average. Computer programs then search for overlap among the sequences and connect and order the fragments into a linear whole. By chance, some stretches will be covered many times over, whereas others will be missed completely; the more sequencing, the fewer gaps will remain.