The objective of this proposal is to significantly improve automated determination of DNA sequences. Practical performance limits of automated DNA sequencers are determined by the separation of oligonucleotides effected by polyacrylamide gel electrophoresis. Designs of contemporary instruments are basically similar. As oligomers in a DNA sequencing ladder pass the detector(s), multi-component analysis specifies the radioactive or fluorescent label associated with each oligomer. Under ideal conditions, determination of the sequence of terminal nucleotides is straightforward. When separations of oligomers or signal levels are not optimal, ambiguities or errors are likely. These are miscalled bases, extra or missing bases, or unidentified bases in the DNA sequence file, typically at about 1 to 3 errors per 100 bases. An error rate near 1% is a common target for DNA sequencing performance, since comparison with complementary strand sequence data should then reduce errors to about 1 per 10,000 base pairs. This is only possible if every mismatch of the sequence and its complement is identified and correctly reconciled. Even then, error rates from 0.01% to 0.1% approximate the variation among alleles in a gene pool: some such alleles can correlate with severe burdens of inherited pathology. Small improvements in single strand error rate will have substantial impact on quality of finished sequences from 1/10,000 bp to 1/1,000,000 bp. Improvements are needed if automated systems are to provide longer spans of DNA sequences with fewer errors. The emphasis of this proposal is on raw data acquisition and new methods for translation of the raw data to finished DNA sequences. An expert system, rule-based method will be developed to reinforce conventional translation of raw data to DNA sequences. An independent, pattern-recognition system will also be developed and tested, using techniques for construction and training of neural nets. We will also evaluate two new approaches to utilize single label, single data channels for more efficient determination of DNA sequences. Alternative approaches to oligonucleotide separation for sequence analysis will also be investigated. In pursuit of these specific aims we will take advantage of the relative separations and intensities of successive oligomers in DNA sequencing ladders, as independent determinants of DNA sequence-specific data stream patterns.
