New approaches to recognizing functional domains in biological sequences. Final report, April 1, 1993--March 31, 1997

The purpose of this project is to develop new approaches and programs for determining the function of DNA domains. This will aid in the understanding of the sequence data obtained through the Human Genome Project. One of the great challenges of that project is to abstract important biological information from the raw sequences that emerge. The efforts have focused on several areas determining the protein coding regions in genomic DNA; recognizing patterns of DNA binding proteins, including nucleosomes, from the sequence using multi-alphabet analyses; better recognition methods for RNA genes and other patterns where structural considerations are important along with sequence; enhancing the ``Sequence Landscape`` approach to pattern recognition and applying it to various problems in domain classification. GeneParser is the program the authors developed to identify optimal classification boundaries in genomic DNA. This was the first approach to combine several types of evidence into the classification and obtain optimal and suboptimal predictions by a Dynamic Programming algorithm. The authors also explored the use of neural networks to obtain the optimal weighting of the different types of evidence