Bisulfite-Sequencing theoryand Quality Control

Felix Kruegerfelix.krueger@babraham.ac.uk

January 2015

2

• 9:15-10:15 Bisulfite-Seq theory and QC• 10:15-10:30 coffee• 10:30-11:30 Mapping and QC practical• 11:30-12:30 Visualising and Exploring talk• 12:30-13:30 Lunch• 13:30-14:00 Methylation tools in SeqMonk• 14:00-15:30 Visualising and Exploring practical• 15:30-15:45 coffee• 15:45-16:35 Differential methylation talk & practical• 16:35-16:45 Introduction to other cytosine

modifications and oxBS

3

DNA Methylation

• Dynamic patterning• Correlation with gene expression• Reset during development• Suppression of repetitive elements • X-chromosome inactivation• Imprinting• Carcinogenesis

Cytosine 5-methyl Cytosine

DNA methyl-transferases

DNA-demethylase(s)?Hydroxymethyl cytosine? TETs?

Passive demethylation?

4

DNA methylation is sequence context dependent

sequenceCCAGTCGCTATACCAGTCGCTATACCAGTCGCTATA

contextCpGCHGCHH

mammalsYESno**no**

plantsYESYESYES

* H being anything but G** somewhat more relevant in certain cell types

5

Promoter methylation causes transcriptional silencing

Adapted from Stephen B BaylinNature Clinical Practice Oncology (2005)

Developmentally important gene or tumour suppressor gene

6

Imprinted Genes: mono-allelic expression

Imprinted Genes: Mono-allelic expression with parent-of-origin specificity. Have key roles in energy metabolism, placenta functions.

Differential allelicDNA methylation X

CGI (CpG island)

methylated CpGunmethylated CpG

7

Imprinted Genes: example

from Ferguson-Smith A. Nat. Rev. Genet. 2011

ICR / DMR

methylated CGI

unmethylated CGI

Rather unusual case where a non-methylated DMR causes silencing of a locus

ICR: Imprinting Control RegionDMR: Differentially Methylated Region

8

DNA methylation is maintained

from W. Reik & J. Walter, Nat. Rev. Genet. 2001

9

DNA methylation is reset during reprogramming

10

Measuring DNA methylation by Bisulfite-sequencing

Image by Illumina

11

Bisulfite Informatics

CCAGTCGCTATAGCGCGATATCGTAme me

TTAGTTGCTATAGTGCGATATTGTA

TTAGTTGCTATAGTGCGATATTGTA

...CCAGTCGCTATAGCGCGATATCGTA...|||||||||||||||||||||||||

Convert

Map

12

- 2 different PCR products and 4 possible different sequence strands from one genomic locus

Bisulfite conversion of a genomic locus

>>CCGGCATGTTTAAACGCT>><<GGCCGTACAAATTTGCGA<<

>>TCGGTATGTTTAAACGTT>><<AGCCATACAAATTTGCAA<<

>>CCAGCATGTTTAAACGCT>><<GGTCGTACAAATTTGCGA<<

Top strand Bottom strand

CTOTOT CTOB

OB

mC|

mC|

|mC

|mC

>>UCGGUATGTTTAAACGUT>><<GGUCGTACAAATTTGCGA<<

Bisulfite conversion

PCR amplification

|hmC

- each of these 4 sequence strands can theoretically exist in any possible conversion state

TTGGCATGTTTAAACGTT

cc..C.........Z.c.

…TTGGTATGTTTAAATGTT……AACCATACAAATTTACAA…

…CCAACATATTTAAACACT……GGTTGTATAAATTTGTGA…

align to bisulfite converted genomes

forward strand C -> T converted genome forward strand G -> A converted genome(equals reverse strand C -> T conversion)

5’…TTGGTATGTTTAAATGTT…3’ 5’…TTAACATATTTAAACATT…3’

(1)

(4)(3)

(2)

sequence of interest

(1) (2) (3) (4)

bisulfite convert read (treat sequence as bothforward and reverse strand)

read all 4 alignment outputs and extractthe unmodified genomic sequence if thesequence could be mapped uniquely

5’…CCGGCATGTTTAAACGCT…3’

CCGGCATGTTTAAACGCTTTGGCATGTTTAAACGTT

methylation call

read sequence

genomic sequence c unmethylated CC methylated Cz unmethylated C in CpG contextZ methylated C in CpG context

Mapping a Bisulfite-Seq read

methylation call

Bismark

14

Common sequencing protocols

>>CCGGCATGTTTAAACGCT>><<GGCCGTACAAATTTGCGA<<

Top strand

Bottom strand

>>TCGGTATGTTTAAACGTT>><<GGTCGTACAAATTTGCGA<<

OTOB

mC|

mC|

|mC

|mC

>>UCGGUATGTTTAAACGUT>>

<<GGUCGTACAAATTTGCGA<<

|hmC

1) Directional libraries (vast majority)

<<AGCCATACAAATTTGCAA<<>>CCAGCATGTTTAAACGCT>>

CTOTCTOB2) PBAT libraries

>>TCGGTATGTTTAAACGTT>><<AGCCATACAAATTTGCAA<<

>>CCAGCATGTTTAAACGCT>><<GGTCGTACAAATTTGCGA<<

CTOTOT

CTOBOB

3) Non-directional libraries

15

Validation

0

20

40

60

80

100

0

20

40

60

80

100

0 20 40 60 80 100

CpG CHG CHH Mapping efficiency

Bisulfite conversion rate (%)

Cyto

sine

sca

lled

unm

ethy

late

d(%

)

Mapping effi

ciency (%)

0

20

40

60

80

100

25 50 75 100 125 150

biasedunbiased

Read length (bp)

Map

ping

effi

cien

cy (%

)

16

BS-Seq Analysis Workflow

QC Trimming Mapping

Methylation extraction

Mapped QCAnalysis

17

Raw Sequence Data

...up to 1,000,000,000 lines per lane

18

Part I: Initial QC -What does QC tell you about your library?

• # of sequences• Basecall qualities• Base composition• Potential contaminants• Expected duplication rate

19

QC Raw data: Sequence Quality

10%

1%

0.1%

Error rate

QC: Base Composition

20

RRBSWGSBS

21

QC: Duplication rate

22

QC: Overrepresented sequences

23

Common problems in BS-SeqP

hred

sco

re

Position in read (bp)

Position in read (bp)

Bas

e co

nten

t (%

)

Not observed in ‘normal’ libraries, e.g. ChIP or RNA-Seq

24

Removing poor quality basecalls

before trimming after trimmingPh

red

scor

e

25

Removing adapter contamination

before trimming after trimming

26

Summary Adapter/Quality Trimming

Important to trim because failure to do so might result in:

Low mapping efficiency

Mis-alignments

Errors in methylation calls since adapters are methylated

Basecall errors tend toward 50% (C:mC)

27

Part II: Sequence alignment –Bismark primary alignment output (BAM file)

HISEQ2000-06:366:C3G4NACXX:3:1101:1316:2067_1:N:0: 99 16 71322125 255 100M = 71322232 207 NTTATTTAGTTTTTTAGGGTTTGTGTGTAGGAGTGTGGGAATTATGTTTTTTATGGTTGATATTTATTTAAAAGTGAGTATAAATTATATATATTTTTTT #1=DDDDDAAFFHIIIA:<FGHCCEFGHD?CFFBBBGEHHGHIII<FEHIIIII==DE??EHHFHEEEEEEEC>;>66;@CDEEEDCEEEEEEEDDDCBB NM:i:14 XX:Z:G8C2C7C21C13C6CC1C17CC3C4CC4 XM:Z:.........h..h.......x.....................h.............x......hh.h.................hh...h....hh.... XR:Z:CT XG:Z:CT XA:Z:1HISEQ2000-06:366:C3G4NACXX:3:1101:1316:2067_1:N:0: 147 16 71322232 255 100M = 71322125 -207 GGTTATTTTATTTAGGGTTATTGTTTTAGAGTTTTATTGTTGTGAACAGATATATGATTAAGGTAATTTTTATAAGGATAATATTTAATTGGAGTTGGTTCCCEEECADCFFFFHHGHGHIIGIHFIJJIJIHFGHGGGEHIJIIJGIGFJJJJJJJJJJIGJJJJGJJJJIIIJJIJIJJJJJJIJHHHHHFFFFFCCC NM:i:21 XX:Z:2G2CC1C1C1C11C11C2C10C1C4CC4C2C1C3C5C2C12C3C1 XM:Z:.....hh.h.h.x...........h...........x..x......X...h.h....hh....h..h.h...h.....h..h............x...h. XR:Z:GA XG:Z:CT XB:Z:1

methylation call

positionchromosome

Read 1

Read 2

sequence

quality

28

Sequence duplication

Complex/diverse library:

Duplicated library:

deduplication10055 17 100 100 71100percent methylation

33 50 100100 100 50 100percent methylation

29

Deduplication - considerations

RRBS

deduplication

Advisable for large genomes and moderate coverage

- unlikely to sequence several genuine copies of the same fragmentamongst >5bn possible fragments with different start sites- maximum coverage with duplication may still be (read length)-fold (even more with paired-end reads)

CCGG

NOT advisable for RRBS or other target enrichment methodswhere higher coverage is either desired or expected

CCGG

30

Methylation extraction

....Z.....h..h.......x.....Z.........x......hh.h.............z....hx...h....hh.Z...

...........hh.h.............z....hx...h....hh.Z...hh....x.....Z.h.....h..h......x...h........

Read 1

Read 2

....Z.....h..h.......x.....Z.........x......hh.h.............z....hx...h....hh.Z...

hh....x.....Z.h.....h..h......x...h........

Read 1

Read 2

CpG methylationoutput

read ID methstate

chr pos context

redundant methylation calls

31

Methylation extraction

CpG methylation output

bedGraph/coverage output

methylationpercentage

methposchr unmeth

32

Part III: Mapped QC -Methylation bias

good opportunity to look at conversion efficiency

33

Artificial methylation calls in paired-end libraries

5’- GGGNNNNNNNNNNNNNNNNNNNNNNNNNNNCCCA -3’3’- ACCCNNNNNNNNNNNNNNNNNNNNNNNNNNNGGG -5’

end repair + A-tailing

34

Specialist applications (I): Reduced representation BS-Seq (RRBS)

Sequence composition bias High duplication rate

35

Fragment size distribution in RRBS

identical (redundant) methylation calls

36

Artificial methylation calls in RRBS libraries

C genomic cytosineC unmethylated cytosine

37

WGBS PBAT

Specialist application (II): Post-bisulfite adapter tagging (PBAT)

suitable for low input material

38

PBAT-Seq

trim off/ ignore first couple of basepairs

39

Bismark workflow

Pre AlignmentFastQC Initial quality controlTrim Galore Adapter/quality trimming using Cutadapt; handles RRBS

and paired-end reads; Trim Galore and RRBS User guideAlignmentBismark Output BAM

Post AlignmentDeduplication optionalMethylation extractor Output individual cytosine methylation calls; optionally

bedGraph or genome-wide cytosine reportM-bias analysis

bismark2report Graphical HTML report generation Example: http://www.bioinformatics.babraham.ac.uk/projects/bismark/PE_report.html

protocol: Quality Control, trimming and alignment of Bisulfite-Seq data

40

Useful links

• FastQC www.bioinformatics.babraham.ac.uk/projects/fastqc/

• Trim Galore www.bioinformatics.babraham.ac.uk/projects/trim_galore/

• Cutadapt https://code.google.com/p/cutadapt/

• Bismark www.bioinformatics.babraham.ac.uk/projects/bismark/

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

• Bowtie 2 http://bowtie-bio.sourceforge.net/bowtie2/

• SeqMonk www.bioinformatics.babraham.ac.uk/projects/seqmonk/

• Cluster Flow www.bioinformatics.babraham.ac.uk/projects/clusterflow/

protocol: Quality control, trimming and alignment of Bisulfite-Seq datahttp://www.epigenesys.eu/en/protocols/bio-informatics/483-quality-control-trimming-and-alignment-of-bisulfite-seq-data-prot-57

41

FastQC: quality control for high throughput sequencing

FastQ Screen: organism andcontamination detection

Hi-C mapping

SeqMonk: Genome browser, quantitation and data analysis

Bismark: Bisulfite-sequencingalignments and methylation calls

Sierra: A web-based LIMS systemfor small sequencing facilities

ASAP: Allele-specific alignments

Trim Galore! Quality and adaptertrimming for (RRBS) libraries

www.bioinformatics.babraham.ac.uk