Fast and Accurate Short Read Alignment with Burrows-Wheeler TransformHeng Li and Richard Durban

CSC2431February 3rd 2010

Alecia Fowler

Short Read Alignment

SPEED AND ACCURACY

Burrows Wheeler Aligner

OVERVIEW: based on backward search and Burrows-Wheeler Transform (BWT)

FEATURES: performs gapped alignment for single-end reads, supports paired-end mapping, generates mapping quality

PLATFORM: Illumina; SOLiD; 454; Sanger PROS: fast CONS: short read algorithm is slow for long reads

and reads with high error rate

Prefix trie

X = GOOGOL$

“G”“GO”“GOO”“GOOG”“GOOGO”“GOOGOL”

Burrows-Wheeler Transform (BWT)

Algorithm used for data compression

Output is easier to compress as it groups similar symbols together

Suffix array interval and sequence alignment

Exact and Inexact Matching

Has to account for mismatches or gaps in the reads

the BWT index of the reverse reference sequence narrows the search space

W = LOL X = GOOGOL$

Evaluation: Simulated Data

Simulated reads from human genome

One million pairs of different lengths

Mapped to the human genome

BWA was found to be more accurate than Bowtie and SOAPv2

Would need to sacrifice mapping quality in order to increase speed

Evaluation: Real Data

12.2 million pairs of 51bp reads from a male genome

Mapped to human genome and a human-chicken hybrid reference

Had high speed and accuracy for both