Algorithms and Computer Architectures for Evolutionary Bioinformatics

Many algorithms in the field of evolutionary Bioinformatics have excessive computational requirements. This holds not only because of the continuous accumulation of molecular sequence data, which is driven by significant advances in wet-lab sequencing technologies, but it is also due to the high computational demands of the employed kernels. This dissertation presents new reconfigurable architectures to accelerate parsimonyand likelihood-based phylogenetic tree reconstruction, as well as effective techniques to speed up the execution of a phylogeny-aware alignment kernel on general-purpose graphics processing units. It also introduces a novel technique to conduct efficient phylogenetic tree searches on alignments with missing data. In addition, a highly optimized software implementation for the ω statistic, which is used to detect complete selective sweeps in population genetic data using linkage-disequilibrium patterns of single nucleotide polymorphisms in multiple sequence alignments, is described and made available.