Whole genome comparison

Kelley CrouseAnd

Greg Matuszek

Objective

• Implement a parallel program for genome and chromosome comparisons

Background

• MUMmer: serial implementation using a suffix tree

• Parallel implementation using a variant of the Smith-Waterman local alignment algorithm.

Disadvantages

• Neither handles larger genomes and chromosomes quickly

• Parallel version hindered by data structure

How we plan to implement

• A suffix tree will be created using one sequence

• The second sequence will be fragmented and sent out to the workers.

• Each worker will compare its fragment against the suffix tree and report back to the farmer with the location(s) of similarity

What is a Suffix Tree?

• The tree represents all suffixes within a given string

• Used to search for a sub-string within a string

• By comparing a test string, T, against the suffix tree of string, S, it is possible to locate any and all possible correlations between the two strings

Suffix Tree - Bananas

• Each suffix of “Bananas” is represented within the suffix tree

• Sub-string S, can be compared to bananas by following the paths of each leaf.

Fragmenting the Second Sequence

Random fragmenting- Difficult to assemble alignment- allows for small and large fragments

Specific length fragments- Restricted to one fragment size- Alignment is easier to assemble

What we hope to gain

• Ability to identify conserved regions between genomes (and chromosomes)

• Conduct comparison between large genomes and chromosomes quickly and accurately