menu

home news biography publication invited talks research members courses software

Research topics.

Network Biology & Comparative Network Analysis

Network Alignment Comparative analysis of biological networks (e.g. protein interaction networks) can provide new insights into the functional organization of cells and reveal the underlying mechanisms of various biological functions. We are working on efficient algorithms that can be used for finding local/global network alignment and network querying.
Accurate & Reliable
Cancer Classification
Pathway-markers and subnetwork-markers have been shown to be more reproducible than gene-markers, and pathway/network-based classifiers often outperform gene-based classifiers in discriminating between different disease phenotypes. We are currently working on building an accurate and reliable cancer classifer based on pathway/subnetwork markers.

Noncoding RNA (ncRNA) Prediction & RNA Sequence Analysis

RNA Structural Alignment RNA structural alignment tries to find the best alignment between RNAs based on their structural similarity as well as their sequence similarity. Profile-csHMMs can be used for finding accurate alignments of RNAs (including pseudoknots), at a relatively low computational cost.
[NOTE] Free software for finding RNA alignments is available at: [PCSHMM]
Fast RNA Search Noncoding RNAs are functional RNA molecules that do no encode proteins. Developing fast and effective tools for finding new ncRNAs is one of the most important problems in computational biology. We investigate various methods that can be used for fast RNA search based on signal processing techniques.
Novel ncRNA Prediction We are currently developing a fast and effective whole genome screening method that can be used to identify novel ncRNA families. The primary goal of this project is to predict new ncRNA families that cannot be easily detected using existing methods due to low sequence conservation.

Probabilistic Models & Algorithms

Context-Sensitive HMM The context-sensitive HMM (csHMM) is an extension of the traditional HMM, where some states have variable emission/transition probabilities that depend on part of the past emissions (called the context). This context-sensitivity increases the descriptive power of the HMM significantly, which can be very useful in various applications.
Profile-csHMM The profile context-sensitive HMM (profile-csHMM) is a subclass of csHMM with a linear repetitive structure. Profile-csHMMs are especially useful in representing RNA families with characteristic secondary structures (including pseudoknots) and building RNA sequence analysis tools.
SCA Algorithm The sequential component adjoining (SCA) algorithm is a generalization of the Viterbi algorithm and the Cocke-Younger-Kasami (CYK) algorithm that can solve the optimal alignment problem of profile-csHMMs. It can be used for analyzing sequences with complicated correlations between distant symbols.


Last update: Aug.11.2009