Bioinformatics Applications

8.7 Summary

The rapid growth of biotechnologies, especially in large sequencing projects, has lead to an explosion of genomic data. For molecular biologists, this mass of data is potentially a rich source of knowledge with appropriate in-silico processes. From a computational point of view, the data being analyzed are primarily text sequences (protein or DNA), and the main task consists of computing similarities. This falls into the string processing computation family which has been studied for a long time, and for which numerous parallel architectures have been proposed. FPGAs are well suited for implementing these regular structures and dedicated reconfigurable accelerators have been designed to speed-up key bioinformatic algorithms. Today, commercial products using FPGAs, derived from previous academic research, exhibit impressive performance compared to parallel microprocessor-based machines.

In this chapter, we have first introduced some important genomic applications to highlight the computer power needed. Then, we focused on the algorithmic problems and their associated hardware implementation:

• the historical dynamic programming algorithm has been detailed together with its systolic structure;

• more recent works on seed-based heuristics to rapidly search genomic banks have been discussed;

• HMMs and language models to perform more complex computations have been presented. Finally, we concluded with some realizations of FPGA accelerators dedicated to genomic computations.