WITH the ink barely dry on the first draft of the human genome, three
companies want to go one stage further. They have announced a $185
million plan to map the identity and function of every protein in the human
body—the human “proteome”.
“It’s our goal to complete the human proteome map in three years,” says Peter
Meldrum, president of Myriad Genetics in Salt Lake City, Utah, the company
leading the project. Hitachi of Tokyo will provide computing hardware for the
task, while software will come from Oracle, the Californian computer giant.
There are worries about a private consortium controlling so much vital
information. Sudhir Saharabudhe, the research chief at Myriad, says the
company’s database of interactions will only be available on subscription,
although there will be a cheaper rate for academics.
Advertisement
But the proteome is much too open-ended for one company to bag the whole lot,
says Ewan Birney of the European Bioinformatics Institute in Cambridge. “As long
as the data gets out there in the end, that’s what’s important,” says Ian
Tomlinson of the Laboratory of Molecular Biology in Cambridge, a founder member
of the Human Proteome Organization, an alliance of institutes and companies that
hopes to identify all human proteins. “The aim is to get drugs and cure people.
So in the end, companies have to be involved.”
There are also doubts about whether the feat can be pulled off
technologically. Tackling the proteome is much harder than the genome. Many
genes code for multiple variants of the same protein. And many proteins are
modified by adding sugar molecules, which play a big role in determining where
proteins go and what they do. What’s more, different proteins can join together
to carry out completely new functions.
Myriad is confident that a new subsidiary, Myriad Proteomics, has the
technology to crack these mysteries. It will be relying on two techniques. The
first, “ProNet”, is an automated version of the so-called yeast two-hybrid
system. A yeast cell engineered to produce a human protein is mated with a
second yeast cell that makes a different human protein. The cells turn blue if
the two human proteins interact. The second technique, “ProSpec”, enables the
company to extract interacting proteins from human cells and identify them using
mass spectrometry.
Last year, however, Stan Fields of Washington University in Seattle showed
that many interactions identified by the yeast system don’t occur naturally.
“This means there could be lots of red herrings,” says Tomlinson.
Saharabudhe says the company has found a way to screen out all but 1 per cent
of these false positives. But he admits that the yeast system is not good for
analysing human proteins that have sugars added, or for proteins that straddle
cell membranes. Yet such proteins carry out many vital functions.
