University of Houston

Department of Computer Science

In partial fulfillment of the Requirements for the Degree of

Master of Science

Jim Wang

will defend his thesis

Signal Extraction Algorithms for Microarray Image Analysis

Abstract

The images generated by microarray experiments can infer large amounts of data in respect to gene expression or mRNA concentration. Analysis of microarray images is a multi-step process including the focus of our work – spot intensity extraction. Many software packages are available using a variety of techniques when converting pixel intensity values to a signal. Noise on these images is extremely common; however most available software do not consider the effects of noise when calculating the signal. A valid signal must truthfully represent the mRNA concentration, not the mRNA and noise concentration.

We developed a Windows-based application called ImageAnalyzer at the University of Houston. Our research is centered on finding a set of metrics to determine the most correct signal value possible to be used by ImageAnalyzer. Since microarray experiments and images can vary, these metrics have adjustable parameters to optimize calculations for the specific image being studied. To test these metrics, we used a liner correlation test to compare the performance of our set of metrics with those most commonly used by other software packages. Also, by selecting a set of experiments under similar conditions, we were able to test the stability of our metrics and their consistency in repetition.

We have found that our metrics do produce better results than the commonly used sum-of-all-pixels intensity metric. The signals from different repetitions show an increasing correlation at an optimized parameter which is different from the value produced by other software packages. Our tests also revealed that the optimal parameter set for one type of array is not necessarily the optimal parameter set for another type of array; thus the flexibility of our set of metrics is both necessary and a great advantage over other solutions.

Date: Wednesday, November 19, 2003

Time: 11:00 AM

Place: 550-PGH

Faculty, students, and the general public are invited.

Thesis Advisor: Dr. Yuriy Fofanov