Low frequency variant detection from FFPE and
cell-free DNA using target capture
Matthew
DNA Technologies, Redwood City, CA, USA;
Introduction
DNA Technologies, Coralville, IA, USA
Results: Variant detection
Quality of FFPE DNA affects library complexity
1.00
2000
200
1000
500
Observed mutant allele frequency
Whole genome library yield (ng)
1500
150
100
50
0
0
0.06
0.15
0.51
Q129/41 Value
5
12
14
Input (ng)
10
25
11
8
12
9
50
7
8
100
6
7
Table 2. Pre-capture PCR amplification cycles performed (by sample type
and amount of starting material).
Library complexity
Library construction
PCR amplification
Starting
material
High
complexity
library
Library construction
PCR amplification
Figure 1. Library construction efficiency affects maximum achievable sequencing
depth. Additional sequencing of the low complexity library will only give more
PCR duplicates. 1.00
0.75
0.50
0.25
Expected mutant allele frequency
Figure 5. Correlation of observed and expected mutant allele
frequencies. Mixtures were created by titrating gDNA from cell line
NA24385 into NA12878 at ratios of 100, 20, 10, 5, 1, 0.5, 0.1, and 0%
(all dilutions performed in triplicate). Variants called for 50 SNP sites.
120
2500
2000
1500
1000
1.00
100
80
60
40
24 ng
20
500
8 ng
8 ng
cfDNA
gDNA
(2-step)
13 ng
0
0
0
25
50
75
100
Input into library construction (ng)
gDNA
(1-step)
High quality Low quality
FFPE
FFPE
Input type
Figure 3. Achievable coverage depth for varying inputs of cell-free DNA, high quality
and low quality FFPE, and genomic DNA. Maximum mean coverage by sample type
and input (left); dotted lines reflect power regression for each sample type. Minimum
input required to reach 500X mean coverage as interpolated using linear regression
analysis of maximum mean coverage (right). All libraries generated using 1-step ligation
unless otherwise noted. Q129/41 = 0.47 and 0.18 for high quality and low quality FFPE
DNA, respectively.
Uniformity of coverage by sample type
0.9
0.6
0.3
0.0
80
100
200
300
Mutant allele
fraction
≥20%
10%
0.50
5%
2.5%
1%
<1%
0.25
0.00
100
200
300
400
500
Median coverage of target (exon) bases
Figure 6. Somatic mutation calling sensitivity as a function of median
target coverage. Variants called on a set of 12 mixed gDNA pools
described in Figure 5.
Conclusions
•  Uniform enrichment of DNA targets combined with efficient
library preparation methods allows for deep coverage of
targeted exons with minimal sequencing data, facilitating
accurate detection of mutations at 1–5% with 500X median
target coverage.
60
40
20
Acknowledgements
0
0
0.75
•  xGen® Lockdown® Probes are compatible with libraries
made from clinically-relevant sample quantities and
qualities.
100
Percent of mean coverage depth
Low
complexity
library
0.25
0.00
Sensitivity (fraction of known mutations called)
Library construction and sequencing: FFPE and gDNA samples
were sheared to an average of 300 bp. All libraries were prepared
with commercial kits using either 1-step or 2‑step adapter ligation,
followed by PCR enrichment (Table 2). Libraries were captured
using the 1197 kb xGen® AML Cancer Panel (Figures 2–4),
or a custom 57 kb panel (Figures 6–7), and sequenced on the
NEXTSeq® Desktop Sequencer (Illumina). Maximum mean
coverage was calculated using the number of unique fragments
present at >75% duplication rate. Variant calling performed using
MuTect v1.1.7.
200
140
Minimum input for 500X (ng)
Table 1. Sample quality assessment for FFPE DNA, high quality cell line
DNA, and sheared cell line DNA. Q129/41 values reflect the ratio of 129 bp
vs. 41 bp amplicon, measured via qPCR.
300
>100 ng
Maximum mean target coverage
Q129/41
>0.4
<0.2
1.4
0.69
Maximum mean coverage by input and sample type
Target bases covered (%)
Sample name
High quality FFPE
Low quality FFPE
Cell line DNA
Cell line DNA sheared to 300 bp
400
0.50
0.00
Figure 2. Quality of FFPE affects library yield and complexity. Whole genome library
yields made from 10 ng of variable quality FFPE DNA using 12 or 16 PCR cycles (left);
dashed line shows recommended input into hybrid capture. Performing additional PCR
amplification cycles prior to enrichment does not affect maximum mean coverage (right).
3000
500
0.06
0.15
0.51
Q129/41 Value
Sample acquisition: Input titration data for high quality gDNA and
cfDNA were generated using the Horizon Discovery TruQ2 and
Multiplex I cfDNA Reference Standard, respectively. FFPE blocks,
procured from Asterand Biosciences, underwent DNA extraction
using the QIAamp® DNA FFPE Tissue Kit. Quality of extracted FFPE
DNA was determined using the KAPA hgDNA Quantification and
QC kit (Table 1). Variant calling was performed on mixed pools of
HapMap DNA sourced from Coriell.
Coverage at
target locus
0.75
Methods
1
14
16
Mirna
2 Integrated
1
Jarosz *
Results
•  Clinical samples are often of poor quality or very limited amounts,
making deep sequencing and the detection of rare mutations
challenging. Variable quality of FFPE-derived DNA presents
additional challenges to reaching 500X unique coverage.
•  Target enrichment strategies that produce highly uniform coverage
require less data to achieve deep sequencing for all regions of
interest.
•  We show that deduplicated mean coverage of at least 500X can
be achieved from as little as 10 ng of high quality gDNA, 25 ng
of FFPE‑derived DNA, or 5–10 ng of cell-free DNA (cfDNA).
Sample type
gDNA/cfDNA
FFPE
2
McNeill ,
Target bases covered at listed depth (%)
1 Integrated
Steve
2
Groenewold ,
Maximum mean target coverage
Madelyn
1
Light ,
0
100
200
300
400
Percent of mean coverage depth
Figure 4. Uniformity of coverage achieved using samples described in Figure 3.
Coverage of non-zero bases up to 300% of the mean coverage depth (left). Mean
cumulative coverage (right); line width corresponds to +/- standard error. Less than
2% of the target space was covered at a depth lower than 20% of the overall mean
(dashed line). Data shown for 5 ng cell-free DNA (n = 12), 10 ng gDNA (n = 3), and
25 ng FFPE (n = 3).
500
We would like to thank Tim
Fennell, Nils Homer, and the
team at Fulcrum Genomics
for variant detection and
bioinformatics guidance. For more information visit www.idtdna.com/xGen
*Corresponding author: [email protected]