Download Rockefeller University Scientist Builds Mutation Analysis Tools with

Customer case study
Field of study: Computational biology
Rockefeller University
Scientist Builds Mutation
Analysis Tools with HGMD
In New York, Yuval Itan is developing new
tools to help scientists sort through exome
and genome data and find disease-causing
mutations more rapidly. HGMD® provided
the foundation for tools designed to improve
gene- and mutation-level data analysis
Sample to Insight
At Rockefeller University in New York, research associate Yuval Itan has
built two computational tools that may fundamentally change how scientists associate genetic mutations with human disease.
As a member of the Casanova lab, Itan puts
exome or genome and efficiently home in on
his computational biology background to use
the mutation of interest.
in the effort to understand the genetic underpinnings of Mendelian disease. The lab focuses
To accomplish this goal, Itan turned to a
primarily on children, tracking genetic muta-
resource he trusts: Human Gene Mutation
tions that confer heightened susceptibility to
Database (HGMD), a carefully annotated col-
diseases such as flu or tuberculosis.
lection of data associated with observed mutations. HGMD contains all known human inher-
Itan and his colleagues use whole exome and
ited disease mutations, allowing scientists to
whole genome sequencing to analyze patients,
determine quickly whether a variant is novel or
mostly with severe cases of infectious disease
has been reported previously and to discover
that are suspected of having a genetic compo-
the genetic basis of a disease.
nent, often comparing those patients to healthy
parents or to other patients with similar pheno-
HGMD proved to be just the repository Itan
types to find relevant mutations. “I’m interested
needed to build the new tools that will be freely
in correlating the specific genotypes of the
available to the genomics community: the gene
patients to the specific phenotypes they are
damage index and the mutation significance
displaying,” he says. “We deal with huge data
cutoff. Both tools provide scientists a stream-
— it’s really a needle-in-the-haystack problem.”
lined workflow to find causal mutations as well
as greater confidence in their results.
In Itan’s view, the scientific problem is just as
much about finding the right needle as it is
about not getting distracted by the rest of the
GDI
haystack. In the quest for a causative mutation,
2
chasing down unrelated mutations is costly,
The gene damage index (GDI) was born from
both in time and resources, he says. His goal
the observation that highly mutated genes
in two recent projects was to develop tools
rarely harbor the kinds of disease-causing
that would significantly improve the accuracy
changes that Itan and his fellow scientists are
of mutation prediction and effectively remove
looking for when they dive into the variant lists
as many false leads as possible. These tools
generated from exome or genome sequencing.
will enable other scientists to sift through the
“Genes that are highly mutated in healthy indi-
massive numbers of mutations revealed in any
viduals are very unlikely to cause the severe
Rockefeller University Scientist Builds Mutation Analysis Tools with HGMD
diseases we’re investigating,” Itan says. “A
by exome or genome sequencing. Using this
big proportion of the mutations we find belong
method “massively increases the chance that
to genes that are naturally very highly mutated
investigators will see the true mutation,” Itan
in the general population. These genes can
says. “In many cases mutations that seem
be removed from the analysis, and that would
very promising are not disease-causing and
remove a lot of the noise.”
can be a waste of time.”
Of course, pulling so many mutations out of
the investigation requires an objective, high-
“
We improved the prediction rate from about 40
percent with current methods to about 95 percent
when the mutations’ population frequency is taken
into consideration.
MSC
confidence, gene-level approach focused on
detecting genes irrelevant to disease (rather
For the other tool, Itan zoomed down from
than methods for detecting relevant genes such
the gene level to the mutation level. Aware
as RVIS and de novo excess) — an approach
that many scientists use tools such as CADD,
that didn’t exist when Itan began looking.
PolyPhen-2, or SIFT to determine whether
He reasoned that a careful analysis of how
a mutation will be deleterious or not, Itan
mutated each gene was would provide the
looked into these widely available tools and
foundation for this approach.
found a disturbing trend: many mutations that
were known to cause disease and that are
Itan used HGMD to help with this gene-level
not extremely rare in the general population
analysis. He began by detailing the accumu-
were predicted to be benign. “It’s very risky
lated mutational damage in each human gene,
to remove mutations from the analysis based
assigning a GDI score. The higher the score,
on these predictions,” he says. “If you remove
the more heavily mutated a gene is in healthy
the true mutation, then your project is dead.”
individuals — and therefore less likely to be
associated with disease causation. Once each
Itan believes the problem lies in a uni-
gene was indexed, Itan had to determine an
form benign/damaging cutoff for all human
appropriate threshold between the potentially
genes, based on a single instance per known
disease-causing and the overly mutated genes.
disease-causing mutation. “When frequency
“I took all the disease-causing genes from
of the variant in the population is taken into
HGMD and looked at the profile of the GDI for
account, then about 60 percent of true dis-
genes that are known to be disease-causing,”
ease-causing mutations would be removed
he says. “Then I set the cutoff above which GDI
from the analysis,” he says. The tools he
would not be disease causing.”
assessed seemed to have a strong correlation
to the rarity of a variant.
Evaluations of the GDI approach show that
“these genes can be very safely removed from
That set Itan on the path to creating the
the analysis of patients,” Itan says, noting that
mutation significance cutoff (MSC), a tool
the method allows for stripping out up to 60
designed to provide similar information to
percent of the thousands of mutations detected
scientists as existing tools but with a more
Rockefeller University Scientist Builds Mutation Analysis Tools with HGMD
3
nuanced understanding of the DNA in quesThat work paid off. By setting gene-specific
who have been on diagnostic
RGI
Ingenuity
Variant Analysis
one tool
tion. “Thereodysseys.
is a big difference
in evolutionary
cutoff isvalues
andthey’ve
incorporating those into
enables patients to obtain
genome
or
exome
used
for
RGI
patients,
and
Haraksingh
says
pressure acting upon different human genes,
prediction scores for mutations, “we improved
sequencing, often supported
by crowdfundthat its
of use makes
the application
but currently
available methods
useease
a fixed
the prediction
rate from aabout 40 percent with
ing. But there’s still a cutoff
lot of to
data
interpretation
natural fithe
forsays.
volunteers
who methods
come fromto other
predict
damaging or benign,”
current
about 95 percent when
required to get useful
backthat
to cutoff
backgrounds.
“It allows
people who
aren’t frequency is taken
Withinformation
his approach,
is far more taithe mutations’
population
patients. Often, the lored,
standard
analysis
experts in cull
genomics
to quickly
ask questions
allowing
him topermore accurately
the
into
consideration,”
Itan says. “Also, you can
formed by sequencingdamaging
centers ismutations
not sufficient
about
the
data
without
having
to
worry
about
from all others.
now be relatively confident that if you remove
for understanding these rare and particularly
putting all the components
together
make the
a variant
in atospecific
gene, there is only a 5
challenging cases.
best
says.that
“Variant
Again, Itan relied on HGMD to
getinformatics
started. Hepipelines,”
percentshe
chance
you have removed the true
Analysis
is
so
intuitive
and
easy
to
use
that
looked up each human gene associated with
disease-causing mutation.”
“As the director of RGI’s
Science
2.0 checked
initia- existing
they’re tools’
able toprepick it up quickly.”
disease
and then
tive, I lead a team ofdiction
12 researchers
the
scores for with
the disease-causing
mutations
For both tools, Itan focused on user-friendliness,
mission of analyzing inthe
genomic
Ultimately,
Haraksingh
hopes that
deep
every
gene.and
The mediscores varied
significantly,
designing
them the
for use
by biologists and bioincal data from our patients
to try
figure
out
analysis
to patients
will “Iyield
meaning
thatto an
appropriate
cutoff she
for can
one provide
formaticians
alike.
really hope the GDI will
what’s going on,” Haraksingh
says.
Ultimately,
progress,
if
not
the
exact
answer,
for
each
case
gene would be completely useless for another
help people to have very clean data and to
the volunteers write angene.
in-depth
research
report
—
be
it
a
new
genetic
lead orthe
connecting
For every gene with at least two known
increase
discoverytherate of genes that are
for each patient that disease-causing
is used to refer the
case
patient
to a specialist
who can
provide and
morewith MSC I hope it will
mutations,
Itan
calculated
a
relevant
to disease,
to outside specialists confidence
or garner knowledge
for
insight.
“It’s
been
extremely
gratifying
to
apply
interval based on the expected
help to remove variants that are truly benign
future research from range
RGI’s global
network
of
my
expertise
in
the
nonprofit
says.mutations that are truly
of predictions and established a safe
and helpsector,”
peopleshe
to infer
scientists.
cutoff value. “I would not have been able to
disease-causing,” he says.
do it without HGMD,” he says. “It was crucial.
There was no other way to do this with such
high quality and large scale.”
QIAGEN Silicon Valley
V
1700 Seaport Blvd. Third Floor
Tel. +1 650 381 5100
Redwood City
ty, CA 94063
Fax. +1 650 381 5190
info@ing
genuityy.com
www.ing
genuityy.com
© QIAGEN 2015. All Rights Reserved.
Impact of Human Variation on Disease
4
Rockefeller University Scientist Builds Mutation Analysis Tools with HGMD