Mathematical tool links 11 genetic variations to diabetes
14 October 2009
Mathematicians at Michigan Technological University have developed
powerful new tools for winnowing out the genes behind some of humanity’s
most intractable diseases.
With one, they can cast back through generations to pinpoint the
genes behind inherited illness. With another, they have isolated 11
variations within genes — called single nucleotide polymorphisms, SNPs
or "snips" — associated with type 2 diabetes.
"With chronic, complex diseases like Parkinson's, diabetes and ALS
[Lou Gehrig's disease], multiple genes are involved," said Qiuying Sha,
an assistant professor of mathematical sciences. "You need a powerful
That test is the Ensemble Learning Approach (ELA), software that can
detect a set of SNPs that jointly have a significant effect on a
With complex inherited conditions, including type 2 diabetes, single
genes may precipitate the disease on their own, while other genes cause
disease when they act together. In the past, finding these gene-gene
combinations has been especially unwieldy, because the calculations
needed to match up suspect genes among the 500,000 or so in the human
genome have been virtually impossible.
ELA sidesteps this problem, first by drastically narrowing the field
of potentially dangerous genes, and second, by applying statistical
methods to determine which SNPs act on their own and which act in
To test their model on real data, Sha’s team analyzed genes from over
1,000 people in the United Kingdom, half with type 2 diabetes and half
without. They identified 11 SNPs that, singly or in pairs, are linked to
the disease with a high degree of probability. 
ELA is used to compare the genetic makeup of unrelated individuals to
sort out disease-related genes. The team has also developed another
approach, which uses a two-stage association test that incorporates
founders' phenotypes, called TTFP, that can examine the genomes of
family members going back generations.
"In the past, researchers have dealt with the nuclear family, parents
and children, but this could go back to grandparents, great-grandparents
... as far back as you want."
The team has published their findings in the European Journal of
Human Genetics. 
Now that they’ve developed the software, the analysis is relatively
simple, says Sha. But getting the genetic data to work on is not. "We
don’t have the data sets yet to work with," she says, clearly
frustrated. "That’s the problem with having no medical school."
Those who do have data sets, however, can use the team’s software to
help find the genes associated with a panoply of illnesses. ELA is
available in Windows and Linux versions .
1. Their work was published in the journal Genetic Epidemiology
available online at
2. An abstract is available in the European Journal of Human
3. ELA is available in Windows and Linux versions  at
www.math.mtu.edu/~shuzhang/software.html and TTFP is available by
Bookmark this page