CELERA, the company which shared the glory for sequencing the human genome
last summer, now faces claims that its data may be riddled with errors.
As many as half the company鈥檚 gene sequences for the fruit fly may contain
mistakes, says Samuel Karlin, a mathematician at Stanford University in
California. His work suggests there might be many errors in other genomes that
have been sequenced using similar techniques, he says.
But others have defended the results. 鈥淚 would take 50 per cent correct as a
compliment, not a criticism,鈥� says Gerry Rubin of the US鈥檚 fly-sequencing
programme at the Howard Hughes Medical Institute in Berkeley. An update next
year should improve its accuracy to 75 per cent, he says.
Advertisement
Rubin was one of the coauthors when the head of Celera, Craig Venter,
published the full sequence of the fly genome in Science in March 2000.
The work was hailed as a vindication of the 鈥減ure shotgun sequencing鈥� technique
the company used, where copies of a creature鈥檚 DNA are broken into little pieces
and sequenced. Computers then assemble the full sequence by looking for
overlapping fragments.
But when Karlin began to use Celera鈥檚 fly genome, he spotted lots of errors.
鈥淢ore than 60 per cent of their sequences were in substantial disagreement [with
known sequences], and this got me a little bit angry,鈥� he says. So he did a
fuller analysis with his colleagues. They hunted through Celera鈥檚 genome for
1049 fly genes whose sequence and function had already been worked out and
carefully checked in experiments.
Although 26 per cent of the genes were exact matches, and a further 29 per
cent accurate to within 1 per cent, the remaining 45 per cent contained a range
of moderate to serious errors鈥攐r were missing altogether. 鈥淲e couldn鈥檛
find a lot of the genes we were working on ourselves,鈥� says Karlin. Other errors
included the omission of entire protein-coding segments of genes, or exons,
while other genes had been 鈥渞ead鈥� from the wrong starting point.
Karlin claims there could be worse errors in both the public and private
versions of the human genome. 鈥淓veryone was rushing,鈥� he says. But the public
project relied on a slower, more accurate version of shotgun
sequencing鈥攁nd Celera ended up including a substantial amount of data from
the public project in its version
(New Scientist, 17 February, p 4).
Meanwhile, Karlin warns researchers to take care if they鈥檙e studying any
鈥渘ew鈥� genes revealed by Celera. 鈥淢y advice is to do whatever part you鈥檙e working
on again,鈥� he says. His analysis reveals the perils of relying too much on
computers, Karlin says. 鈥淧eople are trying to get away without doing
别虫辫别谤颈尘别苍迟蝉.鈥�
Celera plays down the criticisms. 鈥淲e never said that the fly genome was
totally and 100 per cent complete,鈥� says spokeswoman Heather Kowalski.
鈥淎nnotation and analyses of genomes can go on for decades.鈥�
Karlin鈥檚 work appears in Nature (vol 411, p 259). He initially sent
his paper to Science but the journal didn鈥檛 publish it. 鈥淭hey sat on it
for five months,鈥� he says.
