DNArm is a project mainly written in C and COMMON LISP, based on the View license.
DNA read mapper will map short reads to the entire genome using all of your cores -- and GPU too
DNArm: A fast DNA read mapper
Project Description:
Develop and implement an algorithm for efficiently mapping short reads of DNA (~30 bases) to the entire genome (~3,000,000,000 bases). In more computer science terms: map strings of length 30 to a constant string of length 3 billion characters. In addition, all mappings that match 28/30 characters should be considered positive.
Parallelization Techniques:
(a) Build a tree of depth 30 with branching factor 5 (one branch to help with fuzzy matching) and have different execution units search through the tree for matches at the same time. The whole tree wont be able to fit into main memory at once but trees are easily broken into components that will fit.
(b) Generate an index of the genome using strings of length 10. Look up each 1/3rd of a short read in the index and then search through the possible locations in different execution units.
(c) Combining the tree, index, and maybe some other types of data structures that we could use to further decompose the search.
Hardware/Technologies: OpenCL
Webpage: http://cs124project-2010.wikidot.com/133-124-joint-proj Google group: http://groups.google.com/group/ucla-cs133cm124