|
November/December
2006
Turning Genotypes into Gold
Many companies founded on genotyping boast
access to special populations and unique data-handling capabilities, but few
have put quite as much muscle into their informatics engines as Genizon
BioSciences.
Genizon's databank is build around whole
genome scans on about 1,000 people from a region that is genetically
relatively isolated. "We've produced a HapMap for Quebec," says
President and CEO John Hooper. Starting with a homogeneous population makes
it easier and cheaper to pick out variants that matter. A study that would
cost about $50 million in a general population can be done for about $5
million in Quebec, according to Hooper.
Based on its proprietary GeneMaps, Genizon
says it has been able to zero in on new disease gene variations for a range
of conditions, including psoriasis, attention deficit hyperactivity
disorder, and baldness. The company does association studies, and
"Almost every time, we've found a disease-related gene we can replicate
in other populations," he says. Underscoring that success, Genizon
recently licensed its Crohn's disease GeneMap to biotech top-gun Genentech.
ASSOCIATION STUDY ENGINE:
Genizon has used its information strength to create a powerful database
for disease gene mining
Besides access to the right kind of genomic
information to start with, "We go for limited amounts of data, but make
sure it is very high quality," says Hooper. "That plays a huge
role in our high success rate."
About 20 people—one-third of the company's
staff—work on data-related issues. Jean Francois Levesque leads that effort.
The company has developed many of its own proprietary algorithms and has
learned when to use certain approaches over others. "We realized right
away that we would have to set up a sophisticated infrastructure and get
professional software engineers involved from the start," he says.
Scientists are interested in what software
does, but not necessarily how it works, hence the need for real developers
too. Having both groups working on the algorithms leads to tools that
"Can deal with 5,000 SNPs as easily as they deal with 50,"
Levesque says. M.A.B.
|
