FPGAs in Bioinformatics : Implementation and Evaluation of
Common Bioinformatics Algorithms in Reconfigurable Logic
- Zugl.: Kiel, Univ. Diss. 2015
Life. Much effort is taken to grant humanity a little insight in this fascinating and complex but fundamental topic. In order to understand the relations and to derive consequences humans have begun to sequence their genomes, i.e. to determine their DNA sequences to infer information, e.g. related to genetic diseases. The process of DNA sequencing as well as subsequent analysis presents a computational challenge for recent computing systems due to the large amounts of data alone.
Runtimes of more than one day for analysis of simple datasets are common, even if the process is already run on a CPU cluster. This thesis shows how this general problem in the area of bioinformatics can be tackled with reconfigurable hardware, especially FPGAs. Three compute intensive problems are highlighted: sequence alignment, SNP interaction analysis and genotype imputation.
In the area of sequence alignment the software BLASTp for protein database searches is exemplarily presented, implemented and evaluated.SNP interaction analysis is presented with three applications performing an exhaustive search for interactions including the corresponding statistical tests: BOOST, iLOCi and the mutual information measurement.
All applications are implemented in FPGA-hardware and evaluated, resulting in an impressive speedup of more than in three orders of magnitude when compared to standard computers. The last topic of genotype imputation presents a two-step process composed of the phasing step and the actual imputation step. The focus lies on the phasing step which is targeted by the SHAPEIT2 application. SHAPEIT2 is discussed with its underlying mathematical methods in detail, and finally implemented and evaluated. A remarkable speedup of 46 is reached here as well.