Vanderbilt Center for Quantitative Sciences Summer Institute Sequencing Analysis (DNA) Yan Guo.

Vanderbilt Center for Quantitative Sciences Summer Institute

Sequencing Analysis (DNA)

Yan Guo

Alignment

ATCGGGAATGCCGTTAACGGTTGGCGT

Reference genome

Human genome is about 3 billion base pair (3,000,000,000)in length.If read is 100 bp long, what is the probability of unique alignment?

1/(4x4x4…4) =1/4100 =1/1.60694E+60

Alignment Tools

• BWA http://bio-bwa.sourceforge.net/

• Bowtie http://bowtie-bio.sourceforge.net/index.shtml

Doing accurate alignment for a 30 million reads will take 30 million x 3billion time units.

Both are based on Borrows-Wheeler Algorithm

Alignment Results – Bam files

• SAM – uncompressed• Bam – compressed• http://

samtools.github.io/hts-specs/SAMv1.pdf• Sort and index before performing analysis• Don’t forget to perform QC on alignment

How to call SNPs

http://www.broadinstitute.org/igv/

Local Realignment

Recalibration

Why do we need realignment and recalibration for DNA but not RNA?

SNP calling

• GATK https://www.broadinstitute.org/gatk/

• Varscan http://varscan.sourceforge.net/

VCF files

Annotation using ANNOVAR

http://www.openbioinformatics.org/annovar/

Somatic Mutation

• Different from SNP (not germline)• Both tumor and normal samples are needed

to accurately define a somatic mutation

• Tumor sample is almost never 100% tumor

Somatic mutation callers

• MuTect http://www.broadinstitute.org/cancer/cga/mutect

• Varscan http://varscan.sourceforge.net/

Quality Control on SNPs

• Number of Novel Non-synonymous SNP ~ 100 – 200

• Transition / transversion ratio• Heterozygous / non reference

homozygous ratio• Heterozygous consistency• Strand Bias• Cycle Bias

Ti/Tv ratio

Heterozygous / non reference homozygous ratio

Ti/Tv ratio by race and regions

Heterozygous / non reference homozygous ratio by race and regions

Heterozygous Genotype Consistency

Strand Bias

Table 1 . Strand bias examples from real data

Chr Pos depth a1 b2 c3 d4

Forward Strand Genotype

Reverse Strand Genotype

6 32975014 21 5 5 10 1Heterzygous Homozygous

1 81967962 38 20 11 7 0Heterzygous Homozygous

12 10215654 31 15 9 7 0Heterzygous Homozygous

1. Forward strand reference allele    

2. Forward strand non reference allele

3. Reverse strand reference allele

4. Reverse strand non reference allele

Cycle Bias

Pooled Analysis

• Pool samples together without barcode• Save money• Can only be used to evaluate allele frequency

Pooled Analysis - Conclusion

Advanced Data Mining

The known and unknown of sequencing data

The known and unknown of sequencing data

KnownUnknown

The known and unknown of sequencing data

KnownKnown UnknownUnkown Unkown

Known – Things we always know that Sequencing data can do

SNV, mutation

CNV

Xie et al. BMC Bioinformatics 2009

Structural VariantsAlkan et al. Nature Review Genetics, 2011

Known Unknown – Other information we found that sequencing data contain

KnownKnown UnknownUnkown Unkown

SNVs and Mutations in non targeted regions Mitochondria

Virus and Microbe

How is additional data mining possible?

• Data mining is possible because capture techniques are not perfect.

Capture Efficiency of The Three Major Capture Kits

Potential Functions of Intron and Intergenic

ENCODE suggested that over 80% human genome maybe functional.

Majority of the GWAS SNPs are not in coding regions (706 exon, 3986 intron, 3323 intergenic)

Coverage of the Unintended Regions

• The coverage don’t just drop off suddenly after the capture region end.

• Capture region example: chr1 1000 1500

1000 1500

1000 1500

Reads Aligned to Non Target Regions Can Be Used to Detect SNPs

• Tibetan exome study : Through exome sequencing of 50 Tibetan subjects, 2 intron SNPs were identified to be associated with high altitude. (Yi, et al. Science 2010)

• Non capture region study: Non capture region’s reads were studied to show they can infer reliable SNPs. (Guo, et al BMC Genomics)

Known unknown - Mitochondria

However, mitochondria is only 16569 BP

Assumptions: 40 mil reads 100BP long read

Dealing with nuMTs

Alignment Results

Extract mitochondria from exome sequencing

Tools:• Picardi et al. Nature Methods 2012• Guo et al. Bioinformatics, 2013 (MitoSeek)

Diagnosis:• Dinwiddie et al. Genmics 2013• Nemeth et al, Brain 2013

Virus

• Virus sequences can be captured through high throughput sequencing of human samples

• HBV in liver cancer samples (Sung, et al. Nature Genetics, 2012) (Jiang, et al. Genome Research, 2012)

• HPV in head and neck cancer (Chen, et al. Bioinformatics, 2012)

HPV AlignmentExample

Tools for Detecting Virus from Sequencing data

• PathSeq (Kostic, et al. Nature, 2011 Biotechnology)

• VirusSeq (Chen, et al. Bioinformatics, 2012)• ViralFusionSeq (Li, et al. Bioinformatics, 2012)• VirusFinder (Wang, et al. PlOS ONE, 2013)

The Data Mining Ideas applied to RNA

• RNAseq has been used a replacement of microarray.

• Other application of RNAseq include dection of alternative splicing, and fusion genes.

• Additional data mining opportunities also available for RNAseq data

SNV and Indel

• Difficulty due to high false positive rate• RNAMapper (Miller, et al. Genome Research,

2013)• SNVQ (Duitama, et al. (BMC Genomics, 2013)• FX (Hong, et al. Bioinformatics, 2012)• OSA (Hu, et al. Binformatics, 2012)

Microsatellite instability

Examples:• Yoon, et al. Genome Research 2013• Zheng, et al. BMC Genomics, 2013

RNA Editing and Allele-specific expression

RNA editing tools and database• DARNED, REDidb, dbRES, RADAR

Allele-specific expression• asSeq (Sun, et al. Biometrics, 2012)• AlleleSeq (Rozowsky, et al. Molecular Systems

Biology, 2011)

Exogenous RNA• Virus (Same as DNA)• Food RNA (you are what you eat)

Wang, et al. PLOS ONE, 2012

nonCoding RNA

Unknown Unknown

Contamination

Reference is not perfect

Unknown treasures

Exome

Samuels, et al. Trends in Genetics, 2013

RNAseq

Quality Control

Quality Quantity

Guo et al. Briefings in Bioinformatics, 2013