New Strategy to detect SNPs
Download as PPT, PDF1 like713 views
This document describes a new strategy for detecting single nucleotide polymorphisms (SNPs) in HIV genetic sequences. The strategy involves trimming low quality bases, correcting base calls using area ratio or average height ratio algorithms, filtering adjacent SNPs, and generating a consensus sequence from multiple alignments. The strategy was tested on 35 batches of HIV sequences and achieved true positive rates of 75.4% using area ratio and 52.6% using average height ratio, outperforming two external SNP detection tools, Polybayes and Polyphred. Future work will involve testing with larger datasets with higher coverage and sequences from more conserved organisms.
New Strategy to detect SNPs
- 1. New Strategy to detect SNPs Miguel Galves José Augusto Quitzau Zanoni Dias Scylla Bioinformatics –Brazil {miguel,jquitzau,zanoni}@scylla.com.br
- 2. Agenda Introduction HIV Dataset Detection Strategy Trimming Procedure Base-Calling Strategies Filter Algorithm Consensus Algorithm Tests Protocol Results Discussion
- 3. Introduction Polymorphism: set of base pair locus at which different alleles exists in individuals in some population – The second most frequent allele must appear in at least 1% of the individuals SNP: polymorphism in a single base pair position SNP discovery is very important to understand complex diseases
- 4. HIV Dataset HIV genetic sequences: – 1302 bp – Well-conserved region 35 batches from 35 individuals: – 6 PCR reads, with average size of 690bp – 1 validated sequence, with manually annotated SNPs HIV Reference Sequence
- 5. Detection Strategy: Survey Trimming Procedure Base-Calling Correction SNPs Filter Batch Consensus Algorithm
- 6. Trimming Procedure Low Quality Ends filtering Converts phred’s quality sequence to error probability sequence: ⇒ Q = -10 x log10(p) Subtract 0.05 from all values (Q=13) Maximum Score Subsequence Algorithm
- 7. Base Calling: Area Ratio The base calling is made in 5 Steps: 1. Chromatogram area delimitation 2. Peak search 3. Choice of the nearest peaks 4. Calculation of the nearest peaks area 5. Calculation of the polymorphic/reference peak area If the calculated ratio is above a certain threshold, the point is considered a polymorphism.
- 8. Base Calling: Area Delimitation
- 9. Base Calling: Peak Identification
- 10. Base Calling: Average Height Ratio Almost the same steps: 1. Chromatogram area delimitation 2. Peak search 3. Choice of the nearest peaks 4. Calculation of the nearest peaks average height 5. Calculation of the polymorphic/reference peak average height. Again, if the calculated ratio is above a certain threshold, the point is considered a polymorphism.
- 11. Base Calling: Peak Identification
- 12. Filter Algorithm Analyzes each sequence Uses a window based algorithm to eliminate adjacents SNPs – Window size: 11 bases – Empirical score system assigned to polymorphism in the window
- 13. Consensus Algorithm Rule-based algorithm – Empirical rules Analyzes the whole cross section to define a consensus – Take account of nucleotide frequencies and qualities Do not create N symbols, nor tri-allelic polymorphisms.
- 14. Consensus Algorithm: Example Sequence 1 A25 C30 C18 C30 A21 Sequence 2 A30 C25 C15 C25 A16 Sequence 3 - M18 A9 C30 - Sequence 4 - - S12 G17 T18 Consensus A M S S W
- 15. Tests Protocol: Third Party Packages Two external packages used to compare our results: – Polybayes: SNP detection tool based on Bayesian Methods – Polyphred: SNP detection tool based on chromatogram analysis ACE file (contig and consensus) created for each batch using phrap ACE file analyzed by Polyphred and Polybayes Results viewed with consed
- 16. Tests Protocol: Our strategy Reads trimmed using Maximum Subsequence Algorithm Base-calling analysis and correction using algorithms describe previously SNP filtering Multiple alignment – Reference sequence as anchor Consensus creation
- 17. Third Party Results: Polybayes Polybayes detected SNPs in only 2 batches out of 35 Batch Existing SNPs Detected SNPs Correct SNPs False Positives False Negatives Batch 13 12 1 1 0 11 Batch 15 5 1 0 1 5
- 18. Third Party Results: Polyphred Polyphred detected SNPs in only 4 batches out of 35 Batch Existing SNPs Detected SNPs Correct SNPs False Positives False Negatives Batch 07 10 1 0 1 10 Batch 14 4 3 0 3 4 Batch 32 26 1 0 1 26 Batch 35 15 8 1 7 14
- 19. Trimming Results Reads average size: – Before trimming: 690.15bp – After trimming: 374.74bp – Reduction of 45% Reference sequence average base coverage – Before trimming: 2.69 – After trimming: 1.77
- 20. Results: True Positive (%) x batch
- 21. Results: False Negative (%) x batch
- 22. Results: False Positive (%) x batch
- 23. Results: Summary Polybayes Polyphred Area Avg. Height Avg SD Avg SD Avg SD Avg SD TP 0.3 1.4 0.2 1.1 75.4 19.2 52.6 21.5 FN 99.7 1.4 99.8 1.1 23.2 18.4 45.6 21.7 DP 0.0 0.0 0.0 0.0 1.4 4.3 1.8 4.0 FP 2.9 16.9 11.1 31.3 393.9 312.3 554.4 511.3 TP + FN + DP = 100%
- 24. Discussion Polybayes and Polyphred need large sets of data to produces good results Our algorithm produces quite satisfactory results taking into account data characteristics: – Low average coverage – High amount of low quality bases – High amount of polymorphisms (virus DNA) Area Ratio strategy produces better results than Average Height strategy
- 25. Future Work Test the algorithms whith larger batches, whith higher average coverage, to improve consensus algorithm Reproduce the experiments using genetic sequences of more conserved life forms, such as mammals
- 26. Acknowledgments