High Sensitivity Sanger Sequencing for Minor Indel Detection and Characterization

Harrison Leong and Stephan Berosik
ThermoFisher Scientific, Genetic Sciences Division, 200 Oyster Point Blvd., South San Francisco, CA, 94080
BACKGROUND
Inserted or deleted genetic material has been associated with many forms
of cancer5,6,7. For disease detection and treatment monitoring it is of
interest to detect low levels of these mutations. To do this with Sanger
sequencing technologies, a complication is that signals from the normal
DNA sequence and mutated DNA are mixed together and the sequences
are shifted relative to one another. Figure 1 shows an example of the
resultant electropherogram caused by the insertion of CGCC in half of the
DNA molecules in the sample.
CONCLUSIONS
In addition to the conclusions expressed in the abstract, a closer look at the results
suggests that the VAF limit for detecting indels may be below 2.5%. The VAF limit
for accurately characterizing indels may be around 10%. Additional work is needed
to confirm these results with independent validation data that includes the 1% and
10% VAF data points. No a priori information about the indels were used by the
algorithm so it is possible that results can be improved for panels of targets where a
priori information is available.
REFERENCES
1.Lin, M.T. et al. (2014), American Journal of Clinical Pathology, June 2014; 141:856-866.
2.Jancik, S. et al. (2012), Journal of Experimental & Clinical Cancer Research 2012; 31:79:1-13.
3.Tsiatis, A.C. et al. (2010), Journal of Molecular Diagnostics, July 2010; 12:4:425-432.
4.Leong, H. et al. (2016), AMP conference poster TT27.
5.Yang H. et al. (2010), BMC Medical Genetics, 11:128.
6.Lengar, P., (2012), Nucleic Acids Research, Vol. 40, No. 14 6401–6413.
7.Ye, K. et al., (2016), Nat Med., January ; 22(1): 97–104. doi:10.1038/nm.4002.
High Sensitivity Sanger Sequencing for Minor Indel Detection and Characterization
(Automated Detection and Characterization of Minor Insertion and Deletion Variants Using Sanger Sequencing)
TT79
MATERIALS AND METHODS
DATA FOR DEVELOPING THE METHOD
The samples used and methods to process those samples is covered in
the abstract. The algorithm requires a quartet of .ab1 files to perform
its analysis: sequencing results for a test and control sample sequenced
in the forward and reverse orientations. The following describes the
quartets used to develop the algorithm:
1130 quartets with no indels present covered VAFs 0%, 0.6125%, 1%,
1.25%, 1.4%, 2%, 2.5%, 5%, 5.28%, 7.5%, 10%, 15%, 20%, 25%, 30%,
and 50%. There were from 0 up to 16 SNPs in these quartets.
Sequence lengths ranged from 125bp to 466bp.
203 quartets had indels and covered VAFs 2.5% (33), 5% (33), 6.25%
(30), 7.5% (1), 12.5% (30), 15% (1), 25% (33), and 50% (33); the
number of quartets for each VAF are in parenthesis. VAF was unknown
for 9 of the quartets. There were 122 deletions from 1bp up to 48bp.
There were 81 insertions from 1bp to 6bp. Sequences lengths ranged
from 114bp to 590bp.
DATA PROCESSING
Data from control and test samples for both forward and reverse
sequencing reactions were processed through the basecaller embedded
within the data collection software that comes with Applied Biosystems™
3500xl Genetic Analyzer. The resulting analyzed data, an example of
which is shown in Figure 3, were processed by the indel algorithm to find
whether or not an indel is present and, if so, to determine indel type,
length, location, sequence, and VAF.
ABSTRACT
(This abstract differs from that submitted to AMP.)
Introduction: Detecting minor genetic variants has become
essential to cancer and infectious disease management. Many
have turned to next generation sequencing to fill this need given
the common misperception that the limit of detection (LOD) for
Sanger sequencing is somewhere between variant allele
frequencies (VAFs) of 15% to 25%1,2,3. Recent developments
have generated algorithmic methods to reduce this limit to 5% for
single nucleotide polymorphisms (SNPs)4. We have invented
algorithms to extend this work to detect and characterize
insertions and deletions (indels). It appears we can detect indels
down to 2.5% VAF. Standard Sanger sequencing protocols can
be used. The method can generate the familiar
electropherogram data display with noise substantially reduced.
Methods: The algorithm utilizes forward and reverse sequencing
reactions of both a control sample and the test sample to detect
whether or not an indel is present, and, if so, characterize the
indel. The algorithm was developed using DNA with indels in 33
different amplicons from 21 different genes and DNA with SNP
variants in 22 different amplicons from eight different genes.
These samples were from DNA reference standards or genomic
DNA or DNA extracted from formalin-fixed, paraffin-embedded
tissues. DNA was quantified using the RNase-P quantitative
polymerase chain reaction assay, and serially diluted. Allelic
proportions spanned 0.6125% to 50% for DNAs with no indels
and 2.5% to 50% for DNAs with indels. Samples were amplified
with AmpliTaq GoldTM 360 Master Mix, cycle sequenced with
BigDyeTM Terminator v3.1, and pre-processed using POP7TM
polymer on the 3500xl Series Genetic Analyzer from Applied
BiosystemsTM. For comparison, the indel data was also
processed using commercially available third party software that
surveys Sanger sequencing data for mutations.
Results: 121 samples had deletions having lengths of 1 to 48
base pairs (bp) and 82 samples had insertions having lengths of
1 to 6-bp. 1130 samples contained 0 to 16 SNP variants. For
indel detection, there were zero false positives and zero false
negatives. For indel characterization accuracy, type was 100%,
length 100%, location 93%, sequence 93%, and VAF 92% within
3% of the expected value. By contrast, the third party software
detected 58% of the indels and characterization accuracy for
type was 56%, length 53%, location 8%, and sequence 37%.
Conclusions: The results suggest the possibility of detecting
the presence of indels down to at least 2.5% using typical
Sanger sequencing protocols. A benefit of the approach is that
existing protocols for visually reviewing the results can continue
to be used and are enhanced because the algorithm generates
results in a familiar form for which the noise has been
substantially diminished. The algorithm may give Sanger
sequencing performance and/or economic advantages in some
molecular diagnostic applications that require finding minor
genetic variants.
Figure 1. Sanger sequencing result of a sample with a CGCC
insertion mutation after base 137 in half of the DNA molecules.
The goal is to disentangle these data to determine the type of indel
(deletion or insertion), its length, its location, its sequence, and the
proportion of DNA molecules having the indel (the variant allele
frequency, VAF). For the case shown in Figure 1 this means concluding
that the two molecules involved have the two sub-sequences shown in
Figure 2 and the one with the insertion has a VAF of 50%:
...AGAGAAGCCCGCC GGCTCTT... (sub sequence of normal DNA)
...AGAGAAGCCCGCCCGCCGGCTCTT... (sub sequence of mutated DNA)
Figure 2. Correct answer for disentangling the data of Figure 1 into
normal and mutant molecules; the insertion is shown in red.
Another complication is that the signals associated with the mutated DNA
may be so small that they are hidden within the noise underlying Sanger
sequencing data, see Figure 3:
Figure 3. A minor indel variant can be hidden in the noise underlying
Sanger sequencing; upper panel shows a 25% VAF variant, lower
panel shows the same variant except at a VAF of 2.5%.
Table 1 reports the accuracy of indel detection and characterization achieved by the
methods described in this communication. Table 2 reports values for the third party
software. False positive rate for the third party software was not measured (n/m)
because the software was set up with parameter values that would maximize its ability
to detect low level variants without regard for false positive detections. For VAF 50%
results, calculated VAF is assigned a value of 50% if it is within 40% to 60%.
Calculated VAF values are used as is for all other VAFs. The VAF tolerance reported
is that required for at least 90% of the samples to fall within that tolerance.
RESULTS
Figure 4 shows electropherograms generated by the algorithm.
Figure 4. The result of processing the data shown in Figure 3.
Note that in the lower panel the peaks of the 2.5% VAF indel
variant are revealed.
Table 1: Indel detection and characterization accuracy, overall and split by VAF
overall 50% VAF 25% VAF 12.5% VAF 6.25% VAF 5% VAF 2.5% VAF
Detection rate: 100% 100% 100% 100% 100% 100% 100%
False positive rate: 0% 0% 0% 0% 0% 0% 0%
Indel type: 100% 100% 100% 100% 100% 100% 100%
Length: 100% 100% 100% 100% 100% 100% 100%
Location, exact:
Within 6 bases:
93%
98%
100%
100%
100%
100%
97%
100%
90%
97%
91%
100%
79%
91%
Sequence: 93% 100% 100% 97% 90% 91% 79%
VAF:
Tolerance:
92%
3%
94%
1%
90%
6%
90%
3%
93%
3%
94%
3%
97%
2%
Table 2: Third party software indel detection and characterization accuracy
overall 50% VAF 25% VAF 12.5% VAF 6.25% VAF 5% VAF 2.5% VAF
Detection rate: 58% 100% 100% 93% 20% 15% 12%
False positive rate: n/m n/m n/m n/m n/m n/m n/m
Indel type: 56% 100% 100% 93% 13% 9% 9%
Length: 53% 97% 100% 90% 10% 3% 6%
Location, exact:
Within 6 bases:
8%
38%
24%
94%
24%
100%
3%
20%
0%
0%
0%
0%
0%
0%
Sequence: 37% 94% 91% 17% 3% 3% 0%
VAF: n/m n/m n/m n/m n/m n/m n/m

High Sensitivity Sanger Sequencing for Minor Indel Detection and Characterization

More Related Content

What's hot (20)

Similar to High Sensitivity Sanger Sequencing for Minor Indel Detection and Characterization (20)

More from Thermo Fisher Scientific (20)

Recently uploaded (20)

High Sensitivity Sanger Sequencing for Minor Indel Detection and Characterization