biocomputing.biolab.cs.kent.edu

Faculty Supervisor: Dr. L. Gwenn Volkert

Our lab is focused on the development of heuristic based tools for
bioinformatics and medical informatics. The students working on these
projects have all demonstrated a keen desire to bridge the
communication gap between biologists and computer scientists. This
often translates into a lot of extra work such as taking courses and
reading papers outside of the traditional computer science domain. As
a group we aim to make this extra work as painless as possible by
ensuring that each project has an external faculty collaborator that
functions minimally as a domain expert to whom we may address
biological or medical questions to. Additionally we have collected a
large repository of useful tutorials and encyclopedia-like reference
materials to which all group members have access.

Current students in our research group are:

Debbie Stoffer
Amin Assareh
Nilgoun Raihani
Ben Hienmann
Shilpa Ramana

Alumni and previous personnel are:

John Gale
Mahesh Tamboli
Katherine Koch
Olena Andriyevska

Our current projects are as follows:

Title: Microarray Expression Analysis
Current Personnel: Available
Faculty Collaborator: Dr. William Lynch

Development of machine learning based research tools for microarray
expression analysis (MEA). MEA is a rapidly growing sub-field of
bioinformatics that entails the design and development of computational
techniques for aiding biologist with the interpretation of the large
amount of data produced in microarray experiments. We are developing
web-based systems that utilize machine learning based algorithms to
effectively reduce the dimensionality of the data. The resulting data
can then be presented to the biologist using a variety of 2D and 3D
visualization tools. When possible we utilize existing visualization
tools but are also looking into useful extensions to existing
visualization tools as well as development of new visualization tools.

Title: Phylogenetic Tree Reconstruction
Current Personnel: Available
Faculty Collaborators: Dr. Andrea Schwarzbach

Development of a phylogenetic tree reconstruction algorithm that
supports different models of evolution for constructing different parts
of the phylologeny. Current computational based phylogenetic tree
reconstruction efforts utilize a single model of evolution even if the
biological evidence indicates that the resulting tree is incorrect.
The first step of this project is to quantify how to detect where the
boundaries for different models exists. This project also entails the
combining evolutionary algorithmic approaches and a quartet puzzeling
approaches to the computation phylogenetic tree reconstruction problem.

Title: Microscopic Organism Image Tracking and Analysis
Current Personnel: Mahesh Tamboli
Faculty Collaborators: Dr. S. Vijayaraghvan

Development of efficient methods to yield accurate, real-time path
tracking of real-world point-based objects in successive images with
global optimization of paths and minimization of errors. Our approach
will be to consider all paths concurrently. Existing approaches often
employ greedy algorithms, which work by accepting one best fit possibly
at the cost of large errors elsewhere. We plan to incorporate game
search techniques like tree pruning and minimax algorithms will
facilitate this advance in this class of algorithms.

Title: Medical Data Mining
Current Personnel: Katherine Koch, Olena Andriyevska
Faculty Collaborators: Dr. Yuri Breitbart, Dr. Stephen Ellis

Development of machine learning based data mining techniques for
analysis of medical disease state data such as that provided by Dr.
Steven Ellis of the Cleveland Clinic Cardiac Center. Our first
approach is the development of an evolutionary algorithm approach to
learning a Bayesian Belief Network (BBN) that represents associative
relationships in the data. Bayesian belief networks are an widely
accepted approach for representing relationships among multivariable
datasets, yet the automated construction of BBNs remains a difficult
problem for real life situations due to the enormous space of possible
BBNs for a given dataset. Evolutionary algorithms (EAs) have been
shown to be an effective method for approaching large search problems
such as this. Our second approach is the development of an
evolutionary algorithm approach to learning the model parameters and
topology of a Hidden Markov Model for representing and studying the
same data.

Title: Extending Pairwise Sequence Alignment
Current Personnnel: Ankur Gupta, Darren Brust
Faculty Collaborators: Dr. John Stalvey

Investigation of an evolutionary algorithm approach to pairwise
sequence alignment. Although pairwise sequence analysis can be
completed using standard techniques such as BLAST, there are situations
when this class of algorithms fails to identify significant alignments
that are know to exist. By utilizing additional information available
in the sequence database record in addition to the sequences themselves
we search in the space of all possibele alignments and evolve
potentially significant alignments using an evolutionary programming
algorithm. Sequence alignment of nucleotides in promoter regions of
eukaryotic DNA presents additional problems that we are approaching by
the development of hueristic based pattern identification and
visualiztion algorithms.