Current Research
(This is not the complete list of our projects.)


Intercellular Cell Dynamics Analysis System :
New discoveries in biology have required an extensive knowledge of cell dynamics. Knowledge of subcellular particles/structures such as organelles vesicles, and mRNAs is critical to understand how cells regulate delivery of specific proteins from the site of synthesis to the site of action for a better understanding of diseases and viral infections.  More


Proteomics Biomarker Information System :
We have used high resolution mass spectrometry (MS), MALDI-TOF (matrix-assisted laser desorption/ionization time-of-flight) and SELDI-TOF (surface-enhanced laser desorptin/ionization time-of-flight), to study serum proteomic profiling. To analyze the complex mass spectra whose number of mass-to-charge data points (a.k.a. features in pattern recognition) easily goes beyond one million, sophisticated bioinformatics algorithms have to be explored to robustly discover and identify these unique proteomic patterns.  More


Marker Gene Selection and Clustering :
PFSBEM (hybrid PCA based Feature Selection and Boost- Expectation-Maximization clustering) is for unsupervised gene selection. It uses a two step approach to select marker genes. The first step retrieves feature subsets with original physical meaning based on their capacities to reproduce sample projections on PCs. The second step then searches for the best feature subsets that maximize clustering performance. To improve the quality of partitioning, we explored cluster ensemble approaches based on boosting and cluster validity index.  More

Pro ID

Protein/Peptide Identification by Tandem Mass Spectrometry :
To find the cause or consequence of pathology, a two-way parallel searching for de novo peptide identification has been developed to greatly reduce the number of candidate sequences. By utilizing properties of b- and y- ions, our algorithm filters out peptide candidates. A mass-to-charge intensity based criterion is then developed for further pruning.  More


Minimum Error Shape Classifier :
We have developed a weighted hidden-Markov model (HMM) classifier using generalized probabilistic descent method (GPD) for minimum error recognition. Different from traditional Maximum Likelihood (ML) methods, in which classification is based on probabilities from independent individual class models as is the case for general HMM methods, our method utilizes information from all classes to minimize classification error. We have tested different datasets by using our approach and classic ML classification with various HMM topologies alongside Fourier descriptor and Zernike moments based Support Vector Machine (SVM) classification.  More