Publications

Teaching Experience

Office Hours

Current Research Topics

Research Grants Received

List of Graduate Students

Contact Info

Home

 


Translational Bioinformatics for Disease Diagnosis and Treatment Planning

Research Abstract

Microarray technology for measuring gene expression values has created significant opportunities for advances in disease diagnosis and individualized treatment planning. However, the random noise introduced by the sample preparation, hybridization, and scanning stages of microarray processing creates significant inaccuracies in the gene expression levels, and hence presents a major barrier in realizing the anticipated advances. Literature presents several methodologies for noise reduction, which can be broadly categorized as: 1) model based approaches for estimation and removal of hybridization noise, 2) approaches using commonly available image denoising tools, and 3) approaches involving the need for control sample(s). In this research, we present a novel methodology for identifying and removing hybridization and scanning noise from microarray images, using a dual tree complex wavelet transform based multiresolution analysis coupled with bivariate shrinkage thresholding. The key features of our methodology include consideration of inherent features and type of noise specific to microarray images, and the ability to work with a single microarray without needing a control. Our methodology is first benchmarked on a fabricated data set that mimics a real microarray probe data set. Thereafter, our methodology is tested on data sets obtained from a number of Affymetrix GeneChipr human genome HG-U133 Plus 2.0 arrays, processed on HCT-116 cell line at the Microarray Core Facility of Moffitt Cancer Center and Research Institute. The results indicate an appreciable improvement in the quality of the microarray data.

Publications

Conference Presentations

  • A Novel Approach for Denoising Gene Expression Values Obtained from Microarrays", Presentation made in the Quality, Statistics, and Reliability (QSR) section at an Invited Session at INFORMS annual meeting at Pittsburgh , PA , November 2006.
  • "A Multiresolution Approach to Microarray Denoising for Efficient Disease Diagnosis and Drug Discovery”, Poster Presentation made during “Engineering Research Day” at University of South Florida , Tampa , Spring 2007.