Most modern attempts to decipher how portions of
genetic code are translated into physical
characteristics are akin to a first-grader trying to
sound out a word letter by letter — or, in this case,
base pair by base pair.
But University of Florida researchers have
developed a computational method that’s more like
reading whole words at a time.
In a world where science’s ability to transcribe an
organism’s genetic code is growing faster every day,
the technique could offer much needed efficiency in
translating the seemingly endless string of characters
into information that can cure disease or create new
crops.
The researchers, from UF’s Institute of Food and
Agricultural Sciences and the UF Genetics Institute,
published their verification of the method in
Wednesday’s PLoS ONE, an online journal produced by
the Public Library of Science.
“We worked very hard to find ways to collect
genetic information,” said Rongling Wu, the project’s
lead researcher and a UF Research Foundation
professor. “We now must work hard to find ways to use
it.”
In many respects, researchers think of an
organism’s genome as ticker-tape listings of four
letters — representing four amino acid bases —
repeated in varying orders. The goal is to find
meaning within the sequences, to figure out how
variations in the pattern affect the organism’s
physiology.
Humans, for example, have 3 billion letters in our
code. Between any two of us, 99.9 percent of those
letters are the same. But it’s that last 0.1 percent
of difference, peppered throughout our DNA in the form
of single-letter changes, that accounts for our unique
identities—from eye color to disease susceptibility.
These differences are called single nucleotide
polymorphisms, or SNPs (pronounced “snips”).
The simplest way to find out how a SNP affects an
organism is to collect a group of organisms that have
different variations of that letter in their genetic
code.
But physical traits are typically affected by
multiple SNPs that interact in sometimes unpredictable
ways — much like the way an “e” at the end of a word
can change its pronunciation.
Fortunately, the rules of genetics say that SNPs
that affect the same trait are generally related to
each other in some way, such as being near each other.
Wu’s model uses these rules in conjunction with
statistical analysis of real data from genetically
mapped organisms. As a result, the model can find
whole groups of SNPs associated with a physical trait.
Just as an understanding of general phonetic
principles allows a reader to sound out a whole word,
this extra knowledge of genetics allows Wu’s model to
find whole pictures of genome/physical correlations.
“The real promise of Wu’s work is that it could
offer the opportunity for a researcher to not spend a
really disheartening amount of time parsing out
individual nucleotides, and move more directly to
doing the type of genetic work that’s going to have a
greater significance,” said Rory Todhunter, a
researcher working with canine genetics at Cornell
University.
In the paper, the researchers verified their model
using genetic and physical information from mice that
was first collected from the Washington University lab
of James Cheverud in the mid-1990s. They then compared
their results with several years’ worth of genetic
analysis.
This validation was important, said Wei Hou, the
first author of the paper and an assistant professor
at UF’s department of epidemiology and health policy
research. But the analysis of modern data will be the
real key to the technique’s importance. For example,
the mouse genetic information used in this paper
featured only a few thousand SNPs. The July 29 issue
of the journal Nature cited more than 8 million SNPs
for the mouse genome.
“This shows how we need to move beyond looking at
genomes SNP by SNP,” Cheverud said. “Imagine the work
that’s ahead of us if we don’t.”
