A Fast Read Alignment Method based on Seed-and-Vote For Next GenerationSequencing

Transcription

1 A Fast Read Alignment Method based on Seed-and-Vote For Next GenerationSequencing Song Liu 1,2, Yi Wang 3, Fei Wang 1,2 * 1 Shanghai Key Lab of Intelligent Information Processing, Shanghai, China. 2 School of Computer Science and Technology, Fudan University, Shanghai, China. 3 School of Life Sciences, Fudan University, Shanghai, China. GIW 2016

2 Background importance Read alignment is usually the first step to do genome sequence analysis and is usually the most time-consuming one methods based on hash table (BLAST,SOAP,MAQ) based on prefix/suffix trie (BWA,Bowtie,SOAP2)

3 Background number A whole genome reads including 304 million pairs of reads (library size is ~200GB) length The length of one read is 150bp time BWA-MEME needs 74 hours to process this data(single thread)

4 Background number A whole genome reads including 304 million pairs of reads(library size is ~200GB) length The length of one read is 150bp time FSVA only needs 13 hours to process this data(single thread)

5 Methods 1 FSVA 2 3

6 Methods Building hash table Figure.1. The process of key calculation Vectorn represents the vector of coordinates of subsequences with the same key n Figure.2. An example of our hash table. The process of key calculation. X " is a 62bit unsigned integer calculated by converting a 31bp subsequence into a binary number. key " is a 32 bit unsigned integer, calculated by X " modulo a large prime number M.

7 Methods Generating seeds and voting Figure.3. The process of generating seeds and voting

8 Methods The final mapping block should have at least two votes If more than one block has the same most count of votes, we choose one randomly If the alignment has more than two mismatches, we do a Smith-Waterman dynamic programming between the read and the extending block If a seed votes more than 450 coordinates, we think this seed is unrepresentative and drop it

9 Evaluation FSVA BWA-MEM (Li, H. (2013). Aligning sequence reads, clone sequences and assembly contigs with BWA-MEM. arxiv preprint arxiv: ) Bowtie2 (Langmead, B., & Salzberg, S. L. (2012). Fast gapped-read alignment with Bowtie 2. Nature Methods, 9(4), ) Subread (Liao, Y., Smyth, G. K., & Shi, W. (2013). The Subread aligner: fast, accurate and scalable read mapping by seed-and-vote. Nucleic acids research,41(10), e108-e108.)

10 Evaluation Simulated data Wgsim As the reads are fetched from the reference, we know the exact coordinate of each individual read. Thus, we can compare the locations predicted by each alignment tool with the real locations to evaluate the accuracy 1 million reads Base error rate: 0.4% SNP mutation rate: 0.09% Indel mutation rate: 0.01% 125bp and 150bp Single-end and pair-end

11 Evaluation Simulated data Table 1.Evaluation on simulated data Program stime(s) sconf(%) serr(%) ptime(s) pconf(%) perr(%) BWA Subread FSVA Bowtie BWA Subread FSVA Bowtie

12 Evaluation 125bp 150bp Fig.5.The variety between unmapped percent and error rate

13 Evaluation Real data Table 2.Library size of the five datasets Dataset1 Dataset2 Dataset3 Dataset4 Dataset5 Library size 201GB 290GB 298GB 284GB 342GB Table 3.Time cost on real data Tool Time1(m) Time2(m) Time3(m) Time4(m) Time5(m) BWA-MEM Subread FSVA Bowtie

14 Evaluation Real data Fig. 6.The Ti/Tv ratio of the variants called by SAMtools using the result of BWA-MEM, Subread, Bowtie2 or FSVA. Fig.7.Number of variants called by Samtools using the result of BWA-MEM, Subread, Bowtie2 and FSVA separately.

15 Evaluation Real data 89.05% is identified from the results of all the four tools and only 4814 variants are called only via FSVA. This may demonstrate the specificity of called variants based on FSVA is best when compared with the other three tools. Fig.8.A Venn Diagram of the number of variants with high quality.

16 Conclusion FSVA is suitable for long reads with a high quality (low base error rate). In this situation, FSVA can get a relatively alignment result in very short time. For some studies, the data is a tsunami and the very accuracy for an individual is not critical, FSVA will be a good choice.

17 THANK YOU