Effective Heuristic Methods for DNA Strand Design

Sets of DNA strands that satisfy combinatorial and thermodynamic properties play an important role in various approaches to biomolecular computations, nano structure design, molecular tagging, and DNA microarrays. The problem of designing such sets of DNA strands appears to be computationally hard. This thesis introduces new algorithms for design of DNA strand sets that satisfy any of several combinatorial and thermodynamic constraints, which aim to maximize desired hybridization between strands and their complements, while minimizing undesired cross-hybridizations. To heuristically search for good strand sets for bio-computing applications, our algorithms use a conflict-driven stochastic local search approach, which is known to be effective in solving comparable search problems. We describe new and improved thermodynamic measures of the quality of strand sets. With respect to these measures of quality, our algorithms consistently find, within reasonable time, sets that are significantly better than previously published sets in the literature. We also present a detailed analysis and selection of heuristics for improving the quality of DNA strand selection criteria with direct applications in microarray probe design.