Názov:Sampling as an approach to sequencing MinION data
Vedúci:Mgr. Tomáą Vinař PhD.
Kµúčové slová:MinION, Hidden Markov model, sampling, DNA, CUDA
Abstrakt:MinION DNA sequencing platform produces long reads with relatively high error rates caused by a noisy translation of current measurements into the DNA bases representation. It is known, that preprocessing the current measurements with Hidden Markov models before the translation can help reduce the error. We use the Hidden Markov model to sample from a posterior probability and the main focus of our work is to accelerate the sampling algorithm by using the parallelization potential of graphic cards. Specifically we adopted the CUDA platform used for parallelization of C algorithms and implemented the paradigms it provides into the key parts of our software. Subsequently we compared the running times of both the accelerated algorithm with a version that does not use CUDA and runs on CPU. The results show conclusively that for inputs occurring in real life applications our algorithm offer a significant speedup, ofter in the order of several magnitudes compared with a non-parallel CPU version.

Súbory bakalárskej práce: