| |
 | |
 | |
 | |
 | |
 | |
 | Login |
 | |
| Overview | |
| |
| Quality Assurance Methods for Processing Microarray Imagery | |
| |
| Using Visualization for Microarray Data Analysis | |
| |
| Visualization of Soybean Production | |
| |
Contacts: Peter Bajcsy, pbajcsy@ncsa.uiuc.edu
Collaborator: Lei Liu, The W. M. Keck Center for Comparative and Functional Genomics
Collaborator: Lei Liu, The W. M. Keck Center for Comparative and Functional Genomics
|
Problem Definition For this project, we developed a set of novel quality assurance (QA) control methods for processing DNA microarray laser scanned imagery. These methods were designed to detect systematic errors in microarray images and remove any unreliable information from further data analysis. The QA methods were applied after each grid cell with a dot was identified inside of a microarray slide. Each grid cell was screened for errors due to small signal-to-noise ratio (SNR), irregularity in shape (topology) of a microarray dot, and a statistical deviation from a class of expected intensity distributions. This text describes three QA methods based on SNR, topology and statistical distribution analysis.  Approach The QA method based on SNR analysis consisted of two steps. First, analyzing the difference between maximum and minimum intensity values inside of a grid cell enabled the marking of grid cells with no signal with respect to background. Second, a SNR value computed over a dot template in red and green image bands was used for grid cell removal with unreliable information (small SNR). This type of screening was able to reveal problems with spills during spotting or errors during the hybridization process. The goal of the QA method based on topology analysis was to eliminate from further analysis those grid cells that contained foreground (signal) regions of other than the expected dot shape. This method was executed without knowledge of the dot radius by analyzing the intensity distribution of two mutually perpendicular projections of a grid cell, e.g., along row and column axes. A grid cell passes the screening test if both projections demonstrate the same bell-shaped distributions with a theoretical model being 2*cos(x)*constant. This type of analysis was intended to screen errors arising from spotting and hybridization. The third QA method was based on statistical probability distribution functions (PDFs) of background and foreground (signal) pixel intensities inside of a grid cell. It was assumed that the separation of background from foreground is known from previous grid cell analyses. The types of PDFs that model background and foreground were estimated by using higher order central moments and a probability distribution plane. A grid cell passed the screening test if both background and foreground estimated PDF types matched the expected PDF types, for instance, a uniform PDF for highly up-regulated foreground pixels (due to saturated dynamic range) and a Gaussian PDF for background pixels (due to laser scan noise). the model was able to identify good sales prospects and predict their purchase patterns with a high degree of accuracy.  |
|
 Results In summary, our research has produced three highly useful quality assurance methods for processing DNA microarray images. Spotting and hybridization were considered to be the two main sources of unwanted errors. To date, ours is the only system we are aware of that includes topology and statistics based methods for screening DNA microarray images. Publications Peter Bajcsy, Zonglin L. Liu and Lei Liu, "Quality Assurance Methods for Processing Microarray Imagery," Poster Proceedings of the 10th Intelligent Systems for Molecular Biology, Edmonton, Alberta, Canada, 3-7 August 2002, pp 98. Download Document (PDF 3.7 MB) |
