UNLOCKING ACCESS TO SEQUENCE INFORMATION
FROM LOW-INPUT HYBRIDIZATION CAPTURE
Cassie Schumacher, Sukhinder Sandhu, Laurie Kurihara, Sergey Chupreta, Jonathan Irish, Julie Laliberte, Justin Lenhart, Jordan
RoseFigura, Tim Harkins, Vladimir Makarov  Swift Biosciences, 58 Parkland Plaza, Suite 100, Ann Arbor, MI 48103, Tel: 734.330.2568
High Quality DNA Input
Abstract
Hybridization capture is an important tool for exploring exomes and other subsets of
the genome in a cost-effective manner. The use of these techniques, however,
comes at the price of requiring a minimum 100 ng of input material. This
requirement makes this technique prohibitive for precious samples, such as clinical
specimens, where this amount is not readily available. Here, we introduce a novel
library preparation technique before four different hybridization capture assays to
see if we could improve the efficiency of the capture from as little as 1 ng input.
Two capture panels targeted a spectrum of base compositions between high AT-/
GC-rich regions of the genome. We found that there was no significant loss in the
percent of bases on target between 100 ng and 10 ng with the new technique, and
lost only about 10% of these bases when dropped to 1 ng. This is significant
compared with the existing method which lost almost half of the coverage at 10 ng
and dropped to 16x lower coverage at 1 ng.
The other two capture panels probed genes with high clinical relevance. There was
no drop in on-target bases between the 100 ng, 10 ng, and 1 ng samples using the
new technique. This was a dramatic finding compared to the current method which
lost some (~10%) coverage at 10 ng and produced 8x less coverage at 1 ng. We
also used the new hybridization capture technique with this panel and FFPE
samples and saw only a negligible loss of performance with these damaged
samples when compared with high quality genomic DNA.
By utilizing this new technique, we have demonstrated that hybridization capture is
now possible from 1 ng of input with minimal loss of quality. This advancement is
critical for enabling the use of clinical samples with hybridization capture.
®
Accel-NGS
2S Hyb
NimbleGen™ SeqCap™ EZ MedExome
INPUT
100 ng
10 ng
1 ng
 Increased library complexity
 Balanced coverage of AT-/
GC-rich regions
Figure 1. Accel-NGS 2S Hyb workflow. Accel-NGS 2S Hyb has 5 steps: Repair I
dephosphorylates the 5’ ends of the DNA, Repair II repairs and polishes DNA ends,
Ligation I adds the P7 adapter to the 3’ terminus, Ligation II adds the P5 adapter
to the 5’ terminus, and PCR amplifies the library for hybridization capture.
Swift
Kapa
Swift
Kapa
Swift
Kapa
%
DUPLICATION
1
6
5
13
26
71
MEAN
BAIT COV.
50X
51X
52X
47X
37X
10X
Table 1. Comparative performance metrics between Accel-NGS 2S Hyb and
Kapa using SeqCap EZ MedExome hybridization capture. Libraries were made
using HapMap DNA NA12878 (Coriell) with both the Swift Accel-NGS 2S Hyb kit
and the Kapa Library Preparation kit, followed by the NimbleGen SeqCap EZ
MedExome Panel.
%
INPUT PANEL
ALIGNED
Pan-Can 96
100 ng
AML
93
Pan-Can 95
10 ng
AML
93
Pan-Can 94
1 ng
AML
92
%
DUP.
1
2
3
3
20
21
MEAN
BAIT COV.
41X
42X
40X
41X
31X
32X
% COV.
≥ 1X
99
99
99
98
99
99
% COV.
≥ 20X
87
90
86
90
79
84
% ON
TARGET
79
70
78
71
75
69
The Pan-Cancer panel is 0.9 Mb; libraries were normalized to 0.6M reads. The AML panel is 1.1 Mb; libraries were normalized to 1.1M reads.
Table 2. Accel-NGS 2S Hyb performance metrics using the xGen Pan-Cancer
and xGen Acute Myeloid Leukemia Cancer Panels for hybridization capture.
Libraries were made using HapMap DNA NA12878 (Coriell).
FFPE DNA Input
xGen Pan-Cancer Panel
SAMPLE
INPUT
TYPE
Frozen
6 hr. Fix
100 ng
24 hr. Fix
48 hr. Fix
Frozen
6 hr. Fix
10 ng
24 hr. Fix
48 hr. Fix
Frozen
6 hr. Fix
1 ng
24 hr. Fix
48 hr. Fix
%
%
MEAN
ALIGNED DUP. BAIT COV.
96
1
42X
96
1
43X
97
1
44X
97
1
45X
96
3
42X
96
5
41X
97
4
42X
97
8
42X
95
18
33X
94
32
26X
95
31
27X
95
44
22X
% COV.
≥ 1X
99
99
99
99
99
99
99
99
99
99
100
99
% COV.
≥ 20X
91
93
93
88
90
92
93
86
85
77
79
53
% ON
TARGET
80
81
82
82
80
80
81
81
77
74
76
73
The Pan-Cancer panel is 0.9 Mb; libraries were normalized to 0.6M reads.
Small Sample Limitations
Reminder: 1 ng of gDNA = 334 chromosomal copies of any locus
Achieving 1% detection of a mutation
→
3 chromosomal copies
Conversion Rates
Table 3. Accel-NGS 2S Hyb performance metrics using the xGen Pan-Cancer
Panel for hybridization capture with FFPE DNA. To examine the effects of
fixation on sequencing data, libraries were made with the Accel-NGS 2S Hyb kit
from fresh-frozen kidney DNA and from the same sample fixed for 6-, 24-, or 48-hr,
then paraffin embedded.
Agilent
INPUT
20 ng
10 ng
SureSelectXT
Q2
Lab Solutions Comprehensive Cancer Panel
SAMPLE %
TYPE
ALIGNED
97
FFPE
96
FFPE
%
MEAN
DUP. BAIT COV.
18
121X
% COV.
≥ 1X
99
% COV. % ON
≥ 20X
TARGET
99
89
30
99
98
96X
88
The QCCP panel is 1.3 Mb; libraries were normalized to 3M reads
Figure 2. Genomic considerations of low input. Conversion rates measure the
amount of input DNA converted to functional NGS library molecules. Factors
influencing library molecule loss include the degree of DNA damage, the source of
this damage, and size selection. Measuring conversion rates can be problematic as
accurately quantifying the DNA as it changes during the NGS workflow is difficult.
Swift Biosciences calculates conversion rates by the following:
[actual nM library yield] / [theoretical maximum library yield (nM)]
A
Table 4. Accel-NGS 2S Hyb performance metrics using the Agilent
SureSelectXT Q2 Solutions Comprehensive Cancer Panel for hybridization
capture with FFPE DNA. The Accel-NGS 2S Hyb kit was used to make libraries of
three different inputs from the FFPE lung tumor sample. Many thanks to at Q2 Lab
Solutions for generation and sequencing of libraries.
© 2016, Swift Biosciences, Inc. The Swift logo is a trademark and Accel-NGS is a registered trademark of Swift Biosciences. NimbleGen and SeqCap are trademarks of Roche NimbleGen, Inc. xGen and Lockdown are registered trademarks of Integrated DNA Technologies, Inc. SureSelectXT is a product of
Agilent Technologies. Chemagic is a trademark of PerkinElmer Inc. Illumina is a registered trademark of Illumina, Inc 16-0649, 02/16
Liquid Biopsy
xGen Pan-Cancer Panel
B
INPUT
10 ng
0.5%
1%
0.5%
1%
1 ng
Figure 3. Limit of detection schematic. To assess the limit of detection of AccelNGS 2S Hyb, DNA samples from two individuals with different ethnic backgrounds
were used to prepare libraries. 10 ng of DNA from one individual with a 0.5% or 1%
spike-in of the DNA from the second individual was used as the input DNA. Once
libraries were prepared, they were hybridized to xGen Pan-Cancer probes and
SNPs were detected within this panel.
High Quality DNA Input
xGen® Lockdown® Panels
 Broad input range:
10 pg-1 µg
 Compatible with cfDNA and
FFPE samples
METHOD
%
EST. LIBRARY
ALIGNED SIZE (M)
93
1125
93
240
93
275
93
97
93
45
90
7
All libraries were normalized to 39M reads. The MedExome Panel is 64 Mb in size.
 Simple with-bead protocol
 Sequential repair steps
enable use of damaged DNA
Limit of Detection
CHR: POS
REF
10: 123239112 G
12: 40762546
T
19: 40901604
A
11: 32410002
T
10: 8116598
G
19: 1228191
T
ALT
A
C
G
C
A
C
ALLELE FREQUENCIES
% A SPIKE-IN
% B SPIKE-IN
0.5
1
0.5
1
0.6%
1.4%
99%
99%
0.7%
0.9%
99%
99%
0.8%
1.2%
99%
98%
99%
99%
99%
99%
99%
99%
0.5%
0.5%
0.7%
1.1%
1.1%
1.1%
Table 5. SNP detection from DNA spike-in experiment using Accel-NGS 2S
Hyb and the xGen Pan-Cancer Panel. To determine the limit of detection for
calling SNPs using Accel-NGS 2S Hyb, 0.5% and 1% of one DNA was spiked into
100 ng of the other DNA of varying ethnic background. Libraries were sequenced to
an average coverage of 8500X.
High Quality DNA
Input
CHR:
POS
REF ALT
2:
44502788 A
C
9:
100190780 A
G
17:
71197748 G
A
2:
75115108 A
G
4:
5749904
T
C
12:
8757481
G
A
ALLELE FREQUENCIES
% A SPIKE-IN
% B SPIKE-IN
0.5
1
1%
2.5
5
10
0.5
1
2.5
5
10
2% 3%
6%
11% 99%
1%
2% 4%
6%
11% 100% 99% 98% 97% 93%
1%
1% 4%
7%
12% 100% 99% 99% 96% 93%
99% 98% 96% 93%
99% 98% 95% 92% 85% 0%
1%
1%
4%
7%
99% 98% 95% 93% 86% 1%
1%
2%
3%
7%
98% 98% 93% 92% 86% 0%
0%
2%
3%
7%
Table 6. SNP detection from DNA spike-in experiment using the AccelAmplicon Sample_ID Panel. To compare the efficiency of calling SNPs using
Accel-NGS 2S Hyb against a PCR-based method, the Accel-Amplicon Sample_ID
panel was used to make libraries with the same A and B samples and spike-ins
listed above. SNPs could again be detected at the 1% level. Additionally, the 2.5%,
5%, and 10% spike-in levels were also examined. Libraries were sequenced to an
average coverage of 200X.
cfDNA Input
CHR: POS
2: 212244718
12: 25361074
12: 25361142
12: 25361646
12: 40688695
12: 115108136
ALLELE FREQUENCIES
A Background
100%
100%
100%
100%
100%
100%
B Background
0%
0%
0%
0%
0%
0%
C Background
0%
0%
0%
0%
0%
0%
1% A into
10 ng B
0.6%
1.6%
1.1%
1.9%
0.5%
0.7%
1% A into
10 ng C
1.0%
1.9%
0.9%
1.6%
1.1%
2.0%
Table 7. SNP detection from cfDNA spike-in experiment to determine limit of
detection for Accel-NGS 2S Hyb. To determine if SNPs present at 1% allele
frequency could be detected, 1% of cfDNA sample (A) with a unique ethnic
background was spiked into two, 10 ng cfDNA samples (B and C) of different ethnic
backgrounds. SNPs at 100% from sample A could be detected around 1% when
those SNPs are not present in B and C backgrounds. Libraries were sequenced to
an average coverage of 8700X.
%
ALIGNED
98
98
98
97
97
97
97
%
DUP.
0.1
0.1
0.1
17
15
16
11
MEAN
BAIT COV.
50X
49X
50X
37X
39X
38X
41X
% COV.
≥ 1X
99
99
99
99
99
99
99
% COV.
≥ 20X
92
95
93
87
88
90
88
% ON
TARGET
87
85
88
85
86
83
86
The Pan-Cancer panel is 0.9 Mb; libraries were normalized to 0.6M reads.
Table 8. Accel-NGS 2S Hyb performance metrics using the xGen Pan-Cancer
Panel for hybridization capture with cfDNA. Using cfDNA extracted using the
Perkin Elmer Chemagic™ 360, libraries were generated from as little as 1 ng and
10 ng cfDNA using the Accel-NGS 2S Hyb kit.
Agilent SureSelectXT Q2 Lab Solutions Comprehensive Cancer Panel
INPUT
20 ng
10 ng
SAMPLE %
TYPE
ALIGNED
97
cfDNA
96
cfDNA
%
DUP.
20
MEAN
BAIT COV.
111X
% COV.
≥ 1X
99
% COV. % ON
≥ 20X
TARGET
98
80
38
95X
99
98
80
The QCCP panel is 1.3Mb; libraries were normalized to 4M reads.
Table 9. Accel-NGS 2S Hyb performance metrics using the Agilent
SureSelectXT Q2 Lab Solutions Comprehensive Cancer Panel for hybridization
capture with cfDNA. The Accel-NGS 2S Hyb kit was used to make libraries of
three different inputs from the cfDNA of a subject with a lung tumor. Many thanks to
at Q2 Lab Solutions for generation and sequencing of libraries.
Conclusions
In this study, we have demonstrated:
• The Accel-NGS 2S Hyb DNA Library Kit produces
high complexity library with few PCR duplicates,
even at low input.
• The Accel-NGS 2S Hyb Kit produces library with
much higher complexity than the competition,
particularly at low input.
• Through its repair steps and sequential adapter
ligations, the Accel-NGS 2S Hyb Kit enables
sequencing of library for hyb starting from low
input without sacrificing data quality.
• The Accel-NGS 2S Hyb Kit is capable of making
high quality library even from damaged FFPE
samples.
• The Accel-NGS 2S Hyb Kit enables hybridization
capture technology from circulating, cell-free
(cfDNA) input.
• The Accel-NGS 2S Hyb Kit enables detection of
variants even at 1%.
We acknowledge Q2 Lab Solutions
for Agilent SureSelectXT library
preparation and sequencing. We
acknowledge Perkin Elmer for
www.swiftbiosci.com
cfDNA extraction.