Research Projects

• Generating semi-structured gene summaries from biomedical literature
  We developed Gene Summarizer, the first that automatically generates gene summaries from biomedical literature, to enable biomedical researchers to quickly digest the literature. The Gene Summarizer is also integrated with the BeeSpace informatics search and navigation systems, and able to interactively summarize any collection.

• Computational comparative genomics
  We focused on efficiently identifying functionally coupled sequence clusters in comparative genomics. The spatial clustering of genes across different genomes has been used to study important problems in comparative genomics, from identification of operons to detection of homologous regions. The problem was formalized to compute max-gap clusters which is sufficiently different and novel in comparison with existing pattern discovery problems in data mining. We developed a pattern generation and verification approach (MCPaGe) for pair-wise genome comparison. We further extended the max-gap clusters model in multiple-genome comparison and developed another efficient algorithm (MCMuSeC) applying the idea of signature-based filtering.

• Genome wide prediction of cis-regulatory DNA motifs and modules
  The activity of genes in a cell is regulated by other genes, and complex networks of such regulatory interactions orchestrates a precise pattern of expression of genes. We built and applied probabilistic models and algorithms that integrate the heterogeneous and noisy sources of genomics data, in order to deduce regulatory interactions of genes. We performed genome-wide studies of cis-regulatory DNA motifs and modules in multifarious organisms, from yeast to fruit fly and honeybee. Undergoing work includes developing probabilistic models and algorithms for cis-regulatory modules prediction by integrating heterogeneous data like DNA sequence, microarray and biomedical literature etc.
   

• Summarizing facets and opinions in weblogs
  We defined the problem of topic-sentiment analysis on Weblogs and propose a novel probabilistic model to capture the mixture of topics and sentiments simultaneously. The proposed Topic-Sentiment Mixture (TSM) model can reveal the latent topical facets in a Weblog collection, the subtopics in the results of an ad hoc query, and their associated sentiments and dynamics. This model can also be applied to any text collections with a mixture of topics and sentiments, thus has many potential applications, such as search result summarization, opinion tracking, and user behavior prediction.