Analysis of Genomic and Proteomic Sequences using DSP Techniques

摘要: Analysis of biological sequences by detecting the hidden periodicities and symbolic patterns has been an active area research since couple decades. The periodic components help locating biologically relevant motifs such as protein coding regions (exons), CpG islands (CGI) hot-spots that characterize various functions. discrete nature prompted many researchers to use digital signal processing (DSP) techniques for their analysis. After mapping numerical sequences, DSP using filters, wavelets, neural networks, filter banks etc. have developed detect recurring in these sequences. This thesis attempts develop effective based solve some important problems sequence Specifically, statistically optimal null filters (SONF), matched networks algorithms are analysis deoxyribonucleic acid (DNA), ribonucleic (RNA) sequences. In first part this study, DNA investigated order identify locations CGIs regions, i.e., exons. SONFs, which known ability efficiently estimate short-duration signals embedded noise combining maximum signal-to-noise ratio least squares optimization criteria, utilized problems. Basis characterizing exons formulated be used SONF technique solving problems. In second RNA analyzed predict secondary structures. For purpose, on 2-dimensional convolution stem loop structure. knowledge thus obtained then presence pseudoknot, leading determination entire Finally, third thesis, predicting structure identifying hot-spots. a two-stage network scheme is developed, whereas approach proposed. Hot-spots proteins exhibit characteristic frequency corresponding function. A basis function SONFs belonging functional group. Extensive experiments performed throughout demonstrate effectiveness validity schemes investigation. performance proposed compared with previously reported results validated databases containing annotations. It shown result superior those existing techniques.