Current methods for high-throughput resequencing of custom targets

Adam GordonNickerson Lab, UW Genome Sciences

WHI Genetics SIG call3/26/14

Targeted sequencing: Why?

• Genome sequencing is decreasing in cost, but still expensive (and slow)

• Genotyping is cheap and quick, but is limited to a subset of human variation

• Targeted sequencing is a middle ground– High-throughput, cost-effective variant typing and

discovery on genomic regions of interest

• 3 approaches in the Nickerson Lab: Custom Capture, Custom Amplicon, Molecular Inversion Probes (MIPs)

Barcodedlibraryprep

Custom Capture -- Method

Pool

wash

captured DNA

streptavidinbead cleanup

elute

Hyb. biotinprobes

Multiplexed sequencing

Custom Capture -- Details

• Price: $150-$250 per sample depending on target size and sample number

• Throughput: 24-96 samples per lane• Strengths:– Highest quality data– ‘oldest’ method: robust protocols/software/data analysis

pipelines• Weaknesses:– Priciest of the 3 methods– Hybridization-based capture can have issues with complex /

repetitive regions

PGRNseq: custom capture of pharmacogenetic targets

• Target: 84 PGx-associated genes (exons + UTRs + 1.5kb upstream/downstream)– ~1 Mb total

• Price: $230 / sample (for ~200 samples)

PlexLevel

Avg. #Reads

(M)

Avg. Unique Aligned

Gb

Avg. Mean Quality Score

Avg. % Q30

Bases

Avg. % Targets

HitAvg.

CoverageAvg. %

Targets at > 20x

Avg. % Targets at > 40x

2x24 16.7 1.37 36.7 92.1 94.7 247x 94.8 93.4

PGRNseq testing: 32 diverse HapMap trios

Custom Amplicon -- Method

Custom Amplicon -- Details

• Price: $100-$200 per sample depending on target size and sample number

• Throughput: 24-96 samples per lane• Strengths:– Cheaper than capture– Extension-ligation could potentially help with sequencing

complex regions• Weaknesses:– Newer than capture: not a lot of data out there yet– Data quality not quite as good as capture– PCR based: some regions cannot be designed

MIPs -- Method

MIPs -- Method

MIPs -- Method

MIPs -- Details• Price: $50-$150 per sample depending on

target size and sample number• Throughput: 48-96+ samples per lane• Strengths:– Highest-throughput of the 3 methods– Cost scales extremely well with sample number– Linked probe design reduces erroneous capture

• Weaknesses:– Design pipelines not as robust– Newer than capture: not a lot of data out there yet

PGx13: MIP-based capture of pharmacogenetic targets

• 13 highest priority, actionable PGx targets– Exons, UTRs, 1.5kb

up/downstream– All variants in these

genes with level 1 evidence (PharmGKB)

– ~100 kB total• Price: $130/sample (for

~200 samples)

Avg. Depth 849x

Avg. Coding Depth 908x

Mean % Target > 20x 91.4

Mean % Coding > 20x 97.6

Mean PGRNseq Concordance (coding only) 99.4%

MeanPGRNseq Concordance(Actionable Variants)

100%

PGx13 testing: 32 diverse HapMap trios

Comparing all 3 methods: same sample, different capture

Custom Amplicon*

Custom Capture

MIPs

*this data from an older version of Illumina’s Custom Amp protocol

max = 189x

max = 2698x

max = 1452x

Some regions are just plain difficult

Sequence homology, repeat content, and %GC can hinder all capture methods

Custom Amplicon*

Custom Capture

MIPs

CYP2D6 Pseudogene Pseudogene

Comparing all 3 methods

• Highest quality data: Capture• Cheapest overall: MIPs• Highest throughput: MIPs

• Custom Amplicon potentially a middle ground, but data from most recent kit is scant