Variant detection
The present invention provides a computer-implemented method for detecting cell-free DNA (cfDNA), such as circulating tumour DNA, in a DNA-containing sample obtained from a patient, the method comprising: (a) providing loci of interest comprising at least 2 mutation-containing loci representative of a tumour of the patient (“patient-specific loci”); (b) providing sequence data comprising sequence reads of a plurality of polynucleotide fragments from a DNA-containing sample from the patient, wherein said sequence reads span said at least 2 mutation-containing loci of step (a); (c) optionally, performing reads collapsing to group the sequence reads into read families; (d) calculating the mutant allele fraction across some or all of said at least 2 patient-specific loci, optionally wherein the mutant allele fraction is calculated by aggregating mutant reads and total reads; (e) classifying the sample as containing or not containing the target cfDNA based on the calculated mutant allele fraction. Also provided a related methods and systems.
1 . A method comprising:
detecting circulating tumour DNA (ctDNA) in a cell-free DNA (cfDNA)-containing sample obtained from a patient, the detecting comprising performing computer-implemented steps of:
(a) providing a list of patient-specific loci of interest comprising at least 2, 3, 4, 5, 6, 7, 8, 9, 10, 50, 100, 500, 1000, 2500 or at least 5000 mutation-containing loci obtained by sequencing a tumour sample of the patient or sequencing DNA obtained from a liquid sample from the patient at a time of high tumour disease burden;
(b) receiving sequence data comprising sequence reads of a plurality of DNA fragments from a cfDNA-containing sample from the patient, wherein said sequence reads span said at least 2, 3, 4, 5, 6, 7, 8, 9, 10, 50, 100, 500, 1000, 2500 or 5000 mutation-containing loci of the list of patient-specific loci of interest;
(c) calculating a mutant allele fraction across all of said at least 2, 3, 4, 5, 6, 7, 8, 9, 10, 50, 100, 500, 1000, 2500 or 5000 patient-specific loci that aggregates across said patient-specific loci: a number of mutant reads at a patient-specific locus and a total number of reads at the patient-specific locus, wherein the mutant allele fraction is calculated by determining a ratio of two quantities: a number of mutant reads aggregated across the patient-specific loci and a total number of reads aggregated across the patient-specific loci;
(d) determining that the mutant allele fraction is statistically significantly greater than a background sequencing error rate;
(e) classifying the tumour or the liquid sample as containing ctDNA based on the determination in step (d); and
(f) treating the patient with an anti-cancer therapy,
wherein the anti-cancer therapy comprises chemotherapy, immunotherapy, and/or radiotherapy.
2 . The method according to claim 1 , wherein the computation of the statistical significance of the mutant allele fraction comprises carrying out a one-sided Fisher's exact test, given a contingency table comprising: the number of mutant reads from the sample, the total number of reads from the sample, and the number of mutant reads expected from the background sequencing error rate.
3 . The method according to claim 2 , wherein the background sequencing error rate has been determined for each class in a plurality of classes of base substitution (“mutation classes”) represented in said at least 2, 3, 4, 5, 6, 7, 8, 9, or 10 patient-specific loci, by trinucleotide context,
and wherein the mutant allele fraction calculation in step (c) is performed for each mutation class, optionally wherein the mutation classes comprise at least 5, 6, 7, 8, 9, 10, 11 or all 12 of the following mutation classes: C>G, G>C, T>G, A>C, C>A, G>T, T>C, A>G, T>A, A>T, C>T and T>C,
and wherein the mutant allele statistical significance computation comprises computing the statistical significance for each mutation class taking into account the background sequencing error rate of that mutation class and combining the computed statistical significance of each mutation class to provide a measure of statistical significance for a global mutant allele fraction of the sample.
4 . The method according to claim 3 , wherein the mutant allele statistical significance computation comprises carrying out multiple one-sided Fisher's exact tests to determine the statistical significance of the number of mutant reads observed given the background sequencing error rate for that mutation class, thereby generating a p-value for each mutation class, and combining the p-values using the Empirical Brown's method to provide a global measure of statistical significance for the mutant allele fraction of the sample.
5 . The method according to claim 1 , wherein the sequence data comprising sequence reads obtained in step (b) represent:
(i) Tailored Panel Sequencing (TAPAS) sequence reads, focussed-exome sequence reads, whole-exome sequence reads or whole-genome sequence reads; or
(ii) sequence reads of a plurality of DNA fragments from a sample obtained from the patient after the patient has begun a course of treatment of the tumour and/or after the patient has had surgical resection of the tumour.
6 . The method according to claim 1 , wherein the patient has, or has had, melanoma, lung cancer, bladder cancer, oesophageal cancer, colorectal cancer, ovarian cancer brain cancer, and/or breast cancer.
7 . The method according to claim 1 , further comprising:
performing reads collapsing to group the sequence reads into read families, wherein performing the reads collapsing comprises grouping reads into read families based on fragment start and end position and at least one molecular barcode,
and wherein a minimum 60%, 70%, 80% or 90% consensus between all family members is required,
and wherein a minimum family size of 2, 3, 4 or 5 is required.
8 . The method according to claim 1 , wherein the analysis of the sample includes a size-selection step which separates out different fragment sizes of DNA.
9 . The method according to claim 8 , wherein the size selection step is carried out prior to sequencing library preparation or after sequencing library preparation, and/or wherein the size selection step is a right-sided size selection employing bead-based capture of gDNA fragments.
10 . The method of claim 8 , wherein the sequence reads are size-selected in silico for reads within the size ranges 115-160 bp, 115-190 bp, 250-400 bp and 440-460 bp in order to enrich for those reads representing ctDNA, or wherein the sample being analysed is subjected to a size-selection step in which genomic DNA (gDNA) fragments of >200 bp, >300 bp, >500 bp, >700 bp, >1000 bp, >1200 bp, >1500 bp or >2000 bp are filtered-out, depleted or removed from the sample prior to analysis, optionally prior to DNA sequencing, to generate a size-selected sample.
11 . The method according to claim 1 , further comprising:
performing reads collapsing to group the sequence reads into read families,
wherein said performing reads collapsing further comprises applying at least one minimal residual disease (MRD) filter selected from the group consisting of:
(i) excluding those loci with >2 mutant molecules; and
(ii) selecting only those fragments which have been sequenced in both forward (F) and reverse (R) direction.
12 . The method according to claim 11 , wherein the mutant allele fraction per locus is weighted by tumour allele fraction or wherein the number of mutant alleles per locus is weighted by tumour fraction.
13 . The method according to claim 12 , wherein the mutant allele fraction per locus is weighted by tumour allele fraction according to the formula:
A
F
c
o
n
t
e
x
t
=
Σ
M
R
D
-
l
i
k
e
l
o
c
i
mutant
families
÷
(
1
-
tumour
AF
)
Σ
M
R
D
-
l
ikeloci
total
families
÷
(
1
-
tumour
AF
)
wherein:
AF context is the allele frequency of a given context; tumourAF is the allele frequency of the locus as determined by sequencing DNA obtained directly from the tumour; and MRD-like loci are the mutation-containing loci determined from the tumour of the patient and to which said MRD filter has subsequently been applied, optionally wherein said context is the trinucleotide context.
14 . The method according to claim 13 , wherein said context is the trinucleotide context and only the 6 trinucleotide contexts having the most significant p-values are combined, optionally wherein the n most significant trinucleotide context p-values are combined according to the formula:
Combined
p
-
value
=
-
2
∑
i
=
1
n
ln
(
p
i
)
,
wherein n=1, 2, 3, 4, 5, 6, 8, 10 or 12.
15 . The method according to claim 12 , wherein the allele fraction is determined according to the formula:
A
F
g
l
o
b
a
l
=
Σ
max
(
A
F
c
o
n
t
e
x
t
-
E
c
o
n
t
ext
,
0
)
×
total
familes
c
o
n
t
e
x
t
Σ
total
families
.
16 . The method of claim 1 , wherein the patient has previously undergone tumour resection, the method further comprising:
(i) sequencing a cfDNA-containing sample that has been obtained from the patient in order to obtain sequence data comprising sequence reads of a plurality of DNA fragments from the sample, wherein said sequence reads span at least 2, 3, 4, 5, 6, 7, 8, 9, 10, 50, 100, 500, 1000, 2500 or at least 5000 loci that have been determined to be mutation-carrying loci in cancer cells of the patient;
(ii) carrying out the detecting of circulating tumour DNA (ctDNA) in a cell-free DNA (cfDNA)-containing sample obtained from the patient using the sequence reads obtained in step (i), wherein the method further comprises determining the recurrence of the cancer in the patient based on at least the classification of the sample in step (d) as containing cfDNA or based on the mutant allele fraction calculated in step (c).
17 . The method according to claim 1 , wherein the sample obtained from the patient is:
(a) a limited volume sample comprising less than one tumour-derived haploid genome or wherein the sequencing data from the sample represents sequencing coverage or depth of less than 1, 2, 3, 4, 10 or 20 haploid genomes; or
(b) a dried blood spot sample; a pin-prick blood sample; an archival blood, serum or plasma sample that is less than 500 μl and that has been stored for greater than 1 day, for at least one month, for at least 1 year, and/or for at least 10 years after collection from the patient; or
(c) a limited volume sample selected from the group consisting of:
(i) a blood, serum or plasma sample of less than 500 μl, less than 400, less than 200, less than 100 μl or less than 75 μl;
(ii) a fine needle aspirate (FNA);
(iii) a lymph node biopsy;
(iv) a urine, cerebrospinal fluid, sputum, bronchial lavage, cervical smear or a cytological sample;
(v) a sample that has been stored for more than 1 year, 2 years, 3 years, 5 years or 10 years from the time of collection from the patient; and
(vi) a sample that has been previously processed and failed quality metrics for DNA or sequencing quality, or a sample that belongs to a set of samples from which other samples have been previously processed and failed quality metrics for DNA or sequencing quality.
18 . The method of claim 1 , wherein the mutant allele fraction is calculated according to the formula:
∑
l
o
c
i
(
mutant
reads
)
patient
-
specific
∑
l
o
c
i
(
total
reads
)
patient
-
specific
wherein (mutant reads) patient-specific is the number of mutant reads at a patient-specific locus and (total reads) patient-specific is the total number of reads at the patient-specific locus.
19 . A method comprising:
detecting circulating tumour DNA (ctDNA) in a cell-free DNA (cfDNA)-containing sample obtained from a patient who has previously undergone treatment for cancer, in order to detect recurrence of said cancer, the detecting comprising performing computer-implemented steps of:
(a) providing a list of patient-specific loci of interest comprising at least 2, 3, 4, 5, 6, 7, 8, 9, 10, 50, 100, 500, 1000, 2500 or at least 5000 mutation-containing loci representative of a tumour of the patient;
(b) providing sequence data comprising sequence reads of a plurality of DNA fragments from a cell free DNA-containing sample from the patient, wherein said sequence reads span said at least 2, 3, 4, 5, 6, 7, 8, 9, 10, 50, 100, 500, 1000, 2500 or 5000 mutation-containing loci of the list of patient-specific loci of interest;
(c) performing reads collapsing to group the sequence reads into read families, wherein said performing read collapsing comprises applying at least one minimum residual disease filter selected from the group consisting of: excluding those loci with >2 mutant molecules, and selecting only those fragments which have been sequenced in both forward and reverse direction;
(d) calculating a mutant allele fraction across all of said at least 2, 3, 4, 5, 6, 7, 8, 9, 10, 50, 100, 500, 1000, 2500 or 5000 patient-specific loci, wherein the mutant allele fraction is calculated by aggregating across said patient-specific loci: a number of mutant reads at a patient-specific locus and a total number of reads at the patient-specific locus, and wherein the mutant allele fraction is calculated by determining a ratio of two quantities: a number of mutant reads aggregated across the patient-specific loci and a total number of reads aggregated across the patient-specific loci;
(e) classifying the sample
(i) as containing ctDNA when the mutant allele fraction is found to be statistically significantly greater than a background sequencing error rate, or
(ii) as not containing ctDNA or having unknown ctDNA status when the mutant allele fraction is not found to be statistically significantly greater than the background sequencing error rate; and
(f) determining the recurrence of the cancer in the patient based at least on the classification of the sample in step (e) as containing ctDNA; and
wherein the patient is determined to have a recurrence of the cancer in step (f) and the method further comprises treating the patient with an anti-cancer therapy,
wherein the anti-cancer therapy comprises chemotherapy, immunotherapy, and/or radiotherapy.