Abstract

The analysis of genetic-sequence data bases is a pattern recognition problem containing several novel features. The constrained nature of images used as inputs can be exploited to achieve quantitative correlation-plane results not possible for more general images. We have developed several symbolic representations of DNA-sequence data that address particular goals of sequence analysis. Important considerations in the design of appropriate representations are discussed. We present the results of optical correlator experiments that employ matched spatial filtering in a Vander Lugt correlator architecture. Experiments include searches for specific substrings within a larger string of sequence data, searches for unspecified substrings based on structural motifs, and searches for substrings in the presence of noise. A potential architecture for a real-time interactive optical correlatoris also examined.

© 1990 Optical Society of America