Genotyping By Sequencing (GBS) Method Overview

Sharon E. MitchellInstitute for Genomic Diversity

Cornell University

http://www.maizegenetics.net/

•Background/Goals

•GBS lab protocol

•Illumina sequencing review

•GBS adapter system

•How GBS differs from RAD

•Modifying GBS for different species

•GBS Workflow

Topics Presented

Background

Genotyping by sequencing (GBS) in any large genome species requires reduction of genome complexity.

II.  Restriction Enzymes (REs)I.  Target enrichment

•Long range PCR of specific genes or genomic subsets

•Molecular inversion probes

•Sequence capture approaches   hybridization‐based (microarrays) 

*Technically less challenging*

• Methylation sensitive REs filter out repetitive genomic fraction

QTL are often located in non‐coding regionsVgt1, Tb, B regulatory regions 60‐150kb from gene

Exon Exon Exon Exon

Exon captureMap large numbers ofgenome‐wide markers

QTL

Create inexpensive, robust multiplex 

sequencing protocol

Low‐ or high‐coverageSequencing Illumina GA

Informatics Pipeline

Anchor markers acrosthe genome

Impute missing dataif needed

Combine genotypic & phenotypic data for QTL mapping, GS and 

GWAS

We have created a public genotyping/informatics platform based on next‐generation sequencing

Open Source

• Method available for anyone to use / modify.• Analysis pipeline details and code are public.• Promote dataset compatibility.• Method published in PLoS ONE to promote accessibility.

• Genotype calls available for public projects.

< 450 bp

Restriction site(   ) sequence tag

Loss of cut site

Sample1

Overview of Genotyping by Sequencing (GBS)

• Focuses NextGen sequencing power to ends of restriction fragments• Both SNPs and presence/absence markers can be scored• Small indels are identified but are not scored

Sample2

• Reduced sample handling• Few PCR & purification steps• No DNA size fractionation• Efficient barcoding system• Simultaneous marker discovery & genotyping

• Scales very well

GBS is a simple, highly multiplexed system for constructing libraries for next‐gen sequencing

GBS 96‐ to 384‐plex Protocol(http://www.maizegenetics.net/gbs‐overview)

1. Plate DNA & adapter pair

2. Digest DNA with RE3. Ligate adapters

GBS Adapters and Enzymes

BarcodeAdapter “Sticky Ends”

Barcode(4‐8 bp)

CommonAdapter

P1 P2

ApeKI G CWGCPstI CTGCA GEcoT22I    ATGCA T

5’ 3’

Restriction Enzymes

Illumina Sequencing Primer 2 

Illumina Sequencing Primer 1 

GBS 96‐ or 384‐plex Protocol(http://www.maizegenetics.net/gbs‐overview)

............. ...

..... ..................... ........... ..

..

. .. ...

.

...

..... ......... .. .........

.....

...

. .. ....... ......

...

.. .......

...

. .

..

.... . .. . ....

1. Plate DNA & adapter pair

5. PCR

Primers

2. Digest DNA with RE3. Ligate adapters

Clean‐up4. Pool samples

. .

..

..

.

....

.. ...

...........

.

.

.. ........

.

.

. ... ...

..

.

.

..

.

.

.

.....

.

.

...... ..

..

.

.

.

.

...

.. ...

.. ...

......

.

..

.....

. ......

....

... ..

...

.. ... .

..

..

.

........

.

. .

.PCR

Pooled Digestion/Ligation Reactions

GBS“Library”

PRC primers:

P1 P2Insert O2O1

GBS 96‐ or 384‐plex Protocol(http://www.maizegenetics.net/gbs‐overview)

............. ...

..... ..................... ........... ..

..

. .. ...

.

...

..... ......... .. .........

.....

...

. .. ....... ......

...

.. .......

...

. .

..

.... . .. . ....

1. Plate DNA & adapter pair

5. PCR

Primers

2. Digest DNA with RE3. Ligate adapters4. Pool DNA aliquots 

6.  Evaluate fragment sizes

Clean‐upClean‐up

Perform Titration to Minimize Adapter Dimers Before Sequencing

Size Standards

LibraryAdapterDimer

1500 bp15 bp

NOTE:  Done once with a small number of samples.Adapter dimers constitute only 0.05% of raw sequence reads

Fluo

rescen

seintensity

Time Time

Optimal adapter amount

Small Fragments are Enriched in GBS Libraries Total Fragm

ents

0%

5%

10%

15%

20%

25%

30%

35%

40%

0 50 100

150

200

250

300

350

400

450

500

550

600

650

700

750

800

850

900

950

1000

1000

0000

REFGENOME

IBM

ApeKI fragment size (bp) >10000

B73 RefGen v1IBM (B73 X Mo17) RILs

0

100000

200000

300000

400000

500000

600000

700000

800000

900000

1000000

384‐plex GBS Results for Maize 

Reads

Mean read count per line = 528,000c.v.  = 0.22

Illumina Sequencing Overview

Flowcell8 channels

Solid Phase Oligos

Illumina Sequencing by Synthesis Review

Based on solid phase‐PCR

Illumina Sequencing Platform Based on:

PCR product attachedto surface

1  Solid‐Phase PCR

Flow cell with bound oligos

Denatured “Library”

Cluster Formation Amplifies Sequencing Signal

Linearization

CBot

Bridge Amplification

PCRCleavage

Illumina Sequencing Platform Based on:

2.  “Sequencing by Synthesis”

ACGTGGC

T

Blocking group promotesconcurrent extension.

ACGTGGC

TG

P1 primer

CA

TTGTGC

Sequencing by Synthesis

FlowcellHiSeq 2000

TGCA

GACGT

GC

T

G

ACGTGGC

T

C

ACGT

GC

TG

G

T         

ACGTGGC

T ACGTGGC

T ACGTGGC

T

N

“In Phase” “Out of Phase”

Read 1

GBS captures barcode and insert DNA sequence in single read

Insert DNA

Barcode

Sequencing Primer 1

Sequencing Primer 2

Illumina

Read 1

Read 2

Sequencing Primer 1

BarcodeRE cut‐site

Insert DNA

GBS

Variable Length GBS Barcodes Solves Sequence Phasing Issues

•First 12 nt used to calculate phasing.•Algorithm assumes random nt distribution.•Incorrect phasing causes incorrect base calls.

Barcode

Illumina

Insert DNA

•Good design and modulating the RE cut site position with variable length barcodes produces even nt distribution. 

Read 2

Read 1

BarcodeRE cut‐site

GBS

Insert DNA

Invariant GBS barcodes cause loss of signal intensity.

Successful GBS sequencing run.

GBS Adapter Design

Barcode Design Considerations• Barcode sets are enzyme specific

– Must not recreate the enzyme recognition site– Must have complementary overhangs

• Sets must be of variable length• Bases must be well balanced at each position• Must different enough from each other to avoid confusion if 

there is a sequencing error.– At least 3 bp differences among barcodes.

• Must not nest within other barcodes• No mononucleotide runs of 3 or more bases

http://www.deenabio.com

Most significant GBS technical issues?

• DNA quality• DNA quantification

Different Ends 50%

Same Ends 50%

100%

Ligation

LigationPCR

GBSStandard

GBS does not use standard “Y” adapters

Denatured “library”

Flow Cell with Bound Oligos

Same‐ended Fragments Do Not Form Clusters

Bridge Amplification

Cluster Formation  &“Linearization”

No P1 bindingsite

Cleaved from surface

< 450 bp

Restriction site(   ) sequence tag

Loss of cut site

Sample1

GBS vs. RAD

• Focuses NextGen sequencing power to ends of restriction fragments• Scores both SNPs and presence/absence markers

Sample2

Digest

Ligate adapters

Pool

Random shear

Size select

Ligate Y adapters

PCR

RADGBS

ReferenceDavey et al. 2011

Modifying GBS

Considerations for using GBS with new species and / or different enzymes.

Why Modify the GBS Protocol?

• More markers• Fewer markers (deeper sequence coverage per locus)

• Increase multiplexing• More genome appropriate (avoid more repetitive DNA classes)

• Other novel applications (i.e., bisulfitesequencing).

Genome Sampling Strategies Vary by Species

Dependent on Factors that Affect Diversity:

•Mating System(Outcrosser, inbreeder, clonal?)

•Ploidy(Haploid, diploid, auto‐ or allopolyploid?)

•Geographical Distribution(Island population, cosmopolitan?)

Other Factors

• Genome size– The size of the genome has some bearing on the size of the fragment pool.

– Amount of repetitive DNA directly correlated with genome size.

• Genome composition– The base composition of the genome can affect the frequency and distribution of the cut sites.

– How repetitive DNA is organized in the genome affects library profiles

Sampling large genomes with methylation‐sensitive restriction enzymes.

5‐base cutter

6‐base cutter

MethylatedDNA

UnmethylatedDNA

GBS LibrarySequenced Fragments

Scrub jay

Vole Giantsquid

Deer Mouse 

TunicateYeast

Solitary Bee

Grape Maize Cacao

Rice

Raspberry

Barley

Cassava

Goose

Sorghum

Cassava

Shrub willow

Optimizing GBS in New Species

MaizeSorghumRiceBarley

SwitchgrassBracypodiumPearl MilletTeosinte

AndropogonFonio

Finger MilletReed Canary Grass

TripsacumSugar caneRye grass

IrisBananaOil Palm

GrapeCassavaCacao

WatermelonAppleHop

EucalyptusCashewAuracariaGuinea YamHorseradishCottonChickpeaLentil  

PineSpruce

Conifers

Flowering Plants

StrawberrySilene

SunflowerSafflowerSoybean

GoldenberryJatrophaWillow

Turnip rapeCanolaPlum

Monkey FlowerOnionPeanutCowpeaSquashTomatoClover

Corn SaladAlmondFlax

PepperCucumberSnap peaGourd

ArabidopsisWillowTea

PotatoCherry

RaspberryPeachKnautiaTomatilloCorn Salad

AlderApricotBean

Pigeon Pea

NeurosporaVerticilliumPhytophthora

Solitary BeeCorn Earworm

Cotton Bollworm

Mexican TetraKillifish

Top MinnowCatfish

Scrub JayGoose

ChickadeeFinches

WeaverbirdAntbirdAntwrenFairy‐wrenGnatwrenManakin

OropendolaSapsuckerWarbler

Woodcreeper

CowPigFox

DolphinSeal

Sea LionWalrus

Water BuffaloCamel

Deer MouseVoleHare

Snow LeopardLynx

NewtFrog

Giant SquidPearl Oyster

Choosing Appropriate Restriction Enzymes:Generalizations from the Bench

ApeKI works well for grasses.

EcoT22I I works well for fish, amphibians and invertebrates.

PstI works well for mammals and birds.

Most frequently asked question for new species:

“How many SNPs will I get”?

Answer: “It depends……”

• Genome size and expected heterozygosityaffects size of fragment pool for desiredamount of sequence coverage (enzyme choiceand multiplex level).

• Amount of extant diversity and how well your samples capture that diversity.

• Reference genome sequence?  3‐4X more SNPs attained by aligning to a reference sequence.

SpeciesGenome Size (Mb) Enzyme Sample Size No. SNPs

Maize 2,600 ApeKI 33,000 2,200K

Rice 400 ApeKI 850 60K

Grape 500 ApeKI 8,000 1,200K

Willow* 460 ApeKI 459 23K

Pine* 16,000 ApeKI 12 63K

Vole* 3,400 PstI 283 53K

Fox* 2,400 EcoT22I 48 16K

Cow 3,000 PstI 48 64K

Neurospora(fungus isolates)

40 ApeKI 384 100K

How many SNPs will I get?

*No reference genome.  UNEAK analysis pipeline used for analysis.  To avoid homology/paralogyissues this pipeline calls SNPs very conservatively.

GBS SNP calls in Sorghum bicolor ‐ Lots of Missing Datars# alleles chrom pos strand RIL_1 RIL_10 RIL_100 RIL_101 RIL_102 RIL_103 RIL_104 RIL_105

S10_13181 T/G 10 13181 + T T T T N T N T

S10_13355 T/C 10 13355 + T T T N T N T T

S10_15605 A/G 10 15605 + N A N A A A A N

S10_15607 A/G 10 15607 + N A N A A A A N

S10_15619 A/G 10 15619 + N A N A A A A N

S10_15629 G/A 10 15629 + G G G G G G G G

S10_15685 C/G 10 15685 + C N C C C N C C

S10_15687 G/A 10 15687 + G N G G G N G G

S10_15699 G/C 10 15699 + G G G G G N G G

S10_15720 A/G 10 15720 + N N N N N N N A

S10_15721 G/C 10 15721 + N N N N N N N G

S10_15722 A/T 10 15722 + N N N N N N N N

S10_15723 T/C 10 15723 + N N N N N N N T

S10_16315 G/T 10 16315 + N G G G G N N G

S10_16419 A/G 10 16419 + A A A A N N A A

S10_16432 C/G 10 16432 + C C C C N N C C

S10_16439 A/G 10 16439 + N N N N N N N N

S10_16497 C/A 10 16497 + C C C N C C N N

S10_16498 A/G 10 16498 + A A A N A A N N

S10_16499 C/A 10 16499 + C C C N C C N N

S10_16573 T/A 10 16573 + T T N N T T T T

S10_17505 G/A 10 17505 + G G N G G G G G

S10_17518 G/T 10 17518 + G N G G G G G G

S10_17528 G/C 10 17528 + G G N G G G G G

S10_17533 C/G 10 17533 + C N C C C C C C

S10_17550 G/C 10 17550 + G N G G G G G G

S10_17551 A/C 10 17551 + A N A A A A A A

S10_17591 G/C 10 17591 + G G G G G G N G

S10_17593 T/C 10 17593 + T T T T T T N T

S10_17613 A/G 10 17613 + A A A A A A N A

Filtering SNPs to remove most of the missing data.

Will be covered later in discussion of TASSEL (http://www.maizegenetics.net/)

Missing Data Strategies

• Impute Missing SNPs.  ‐Many algorithms for doing this.

• Technical Options– Reduce the multiplexing level– Sequence the same library multiple times

• Molecular Options– Choose less frequently cutting enzymes

DNA sample info entered to webform

HTS databaseProject approved?Samples shipped

GBS libraries made

Reference genome

Non‐reference genome

HapMap FileSNPs/SampleCoordinates

Lab Analysis Pipelines

DNA sequence data HapMap File

SNPs/Sample

GBS workflow in the IGD

BioinformaticsKatie HymaJeff Glaubitz

Qi Sun

Method DevelopmentRob ElshireEd Buckler

Sharon Mitchell

GBS Team

Laboratory/ProductionCharlotte Acharya

Wenyan ZhuLisa Blanchard

Jennifer Hagadone

Workshop CoordinatorTheresa Fulton