Thursday Jan 30th, 2008
CSE Room 305
12:00 - 1:00 PM
|
Inferring Progression Models for CGH Data
|
|
Jun Liu |
|
Large scale evolutionary analysis across heterogeneous cancer
types is
one of the most critical tasks in cancer research. One of the
mutational processes that can be monitored genome-wide is the
occurrence of regional DNA Copy Number Alterations (CNAs),
which
may lead to deletion or over-expression of tumor suppressors or
oncogenes, respectively. Recently, the descriptive analysis of
oncogenomic summary data was able to point towards a concordance
of
imbalance profiles from entities of similar histological
categories. However, the analysis of average imbalance profiles
can
not capture the diversity of aberration complexity in the
different
entities. There is a great need for efficient methods that can
accurately model the progression of the cancer markers and
reconstruct
evolutionary relationship between multiple types of cancers.
We have developed an automatic method to infer a graph model for
the
markers of multiple cancers from a large population of
Comparative
Genomic Hybridization (CGH) data. Our method identifies highly
correlated markers across different cancer types. It then builds
a
directed acyclic graph that shows the evolutionary history of
these
markers based on how common each marker is in different cancer
types.
We demonstrated the use of this model in determining the
importance of
markers in cancer evolution. We have also developed a new
method to
measure the evolutionary distance between different cancers
based on
their markers. This method employs the graph model we developed
for
the individual markers to measure the distance between pairs of
cancers. We used this measure to create an evolutionary tree
for
multiple cancers.
Our experiments on Progenetix database show that our markers
are
largely consistent to the reported hot-spot imbalances and most
frequent imbalances. The results show that our distance measure
can
accurately reconstruct the evolutionary relationship between
multiple
cancer types.
|
For
upcoming talks, visit http://www.cise.ufl.edu/dbcenter/seminar.shtml