One of the technologies in DNA mapping is the preparation of electrophoretic gels, which when stained and imaged electronically contain a number of diffuse rectangular objects, called "bands", several pixels in area, which must be detected and localized. This paper describes an algorithm which is central to the automated analysis of such gels, part of a large software product now in use for mapping projects at several genomics laboratories. The problem addressed is that of detecting multiple overlapping pulses of known shape and amplitude. The basic algorithm is one of analysis-by-synthesis, where pulses are added one by one to a forward synthesis model, with revisit to all pulses after each new pulse is added. Because the amplitudes are known there is no risk of significantly overfitting the data with additional model terms. The cost function utilized is a weighted least-squares function, which gives more weight to the shape of the test signal and less to its overall amplitude. Application of this algorithm in the larger context of DNA fingerprinting is discussed.
[1]
Richard Durbin,et al.
Image analysis of restriction enzyme fingerprint autoradiograms
,
1989,
Comput. Appl. Biosci..
[2]
Rodger Staden,et al.
Software for genome mapping by fingerprinting techniques
,
1988,
Comput. Appl. Biosci..
[3]
Marco Marra,et al.
A map for sequence analysis of the Arabidopsis thaliana genome
,
1999,
Nature Genetics.
[4]
Parvaneh Saeedi,et al.
Software for automated analysis of DNA fingerprinting gels.
,
2003,
Genome research.
[5]
R. Wilson,et al.
High throughput fingerprint analysis of large-insert clones.
,
1997,
Genome research.
[6]
Joseph A. O'Sullivan,et al.
Information-Theoretic Image Formation
,
1998,
IEEE Trans. Inf. Theory.
[7]
D. Haussler,et al.
A physical map of the human genome
,
2001,
Nature.