Plant & Animal Genomes XIII Conference

January 15-19, 2005
Town & Country Convention Center
San Diego, CA

Discrimination Of Multiple-Band Peaks In Capillary Fingerprinting Chromatograms

Frank M. You¹ , Ming-cheng Luo¹ , Yong Gu² , Carolin Thomas¹ , Gerard R. Lazo² , Patrick E. McGuire³ , Jan Dvorak¹ , Olin D. Anderson²

¹  Department of Agronomy and Range Science, University of California, Davis, CA 95616, USA

²  Western Regional Research Center, Agricultural Research Service, US Department of Agriculture, 800 Buchanan Street, Albany, CA 94710, USA

³  Genetic Resources Conservation Program, University of California, Davis CA 95616, USA

In many cases of capillary electrophoresis, two or more different fragments of same or similar sizes migrate at the same rate and may generate a single large peak, which will result in bias-estimation of total bands in fingerprints. To discriminate the peak type (single-, double- and triple-band peak) in the processing of fingerprinting data, a complex mathematical algorithm was developed employing both quadratic discriminant function and logistic regression model of peak type against four predictors (relative peak height, relative peak area, peak height ratio and peak area ratio) that are derived from peak height and area. A total of 61,290 peaks from 929 fingerprints of two sequenced wheat BAC clones were used as a training data set. With using this algorithm, 99.85% of single-band peaks were correctly classified into single bands, 94.19 % of double-band peaks into double bands, and 99.51% of triple-band peaks into triple bands. This algorithm has been integrated into GenoProfiler, a cross-platform software package using Java (J2SDK 1.4) for fully automated processing of high-throughput capillary fingerprint data. This software is available upon request via http://wheat.pw.usda.gov/PhysicalMapping/tools/genoprofiler/genoprofiler.html.