Cell-free DNA methylation patterns for disease and condition analysis
View Patent ↗Disclosed herein are methods and systems of utilizing sequencing reads for detecting and quantifying the presence of a tissue type or a disease type in cell-free DNA prepared from blood samples.
1. A method for treating a subject with a cancer, comprising:
identifying a plurality of CpG cluster features of a cancer type t, wherein the plurality of CpG cluster features has a total number K of CpG cluster features, K being a positive integer;
obtaining a cell-free DNA (cfDNA) methylation profile of the subject;
estimating a methylation level x k for each of the plurality of CpG cluster features in the cfDNA methylation profile of the subject, wherein k=1, 2, . . . K;
determining a circulating tumor DNA (ctDNA) burden coefficient θ in the cfDNA methylation profile of the subject, wherein 0≤θ≤1;
calculating a prediction score λ using the ctDNA burden coefficient θ, the cancer type t, and the plurality of methylation levels x k ;
determining that the subject has the cancer type t, when the prediction score λ is greater than a pre-determined threshold; and
administering a treatment to the subject based on the determining that the subject has the cancer type t, wherein the treatment comprises a member selected from the group consisting of a chemotherapy, a radiation therapy, an immunotherapy, and a tumor resection.
2. The method of claim 1 , wherein identifying the plurality of CpG cluster features further comprises:
determining a first methylation level of a CpG cluster in a non-cancerous methylation profile;
determining a second methylation level of the CpG cluster in a cancerous methylation profile;
determining a methylation range coefficient MR based at least in part on a difference between the first methylation level and the second methylation level; and
identifying the CpG cluster as a CpG cluster feature, if the methylation range coefficient MR is greater than a pre-determined MR threshold.
3. The method of claim 1 , wherein estimating the methylation level x k for each of the plurality of CpG cluster features further comprises:
(i) estimating a methylation level x k of a specific CpG cluster using a number of methylated cytosines m k and a total number of cytosines n k within the specific CpG cluster, or
(ii) deriving a probability density function describing the methylation level x k at least in part by convolving a first beta distribution function of a non-cancerous methylation profile and a second beta distribution function of a cancerous methylation profile.
4. The method of claim 1 , wherein the ctDNA burden coefficient θ is determined using a global optimization estimation, wherein the global optimization estimation comprises minimizing an error value according to an error function evaluated at a plurality of different pre-defined θ values.
5. The method of claim 1 , wherein the cancer type t is determined using a global optimization estimation, wherein the global optimization estimation comprises minimizing an error value according to an error function evaluated at a plurality of different pre-defined t values.
6. The method of claim 1 , wherein the prediction score λ is determined using a maximum likelihood estimation, based at least in part on the ctDNA burden coefficient θ, the cancer type t, and the plurality of methylation levels x k .
7. The method of claim 1 , wherein the treatment comprises the chemotherapy.
8. The method of claim 7 , wherein the chemotherapy is selected from the group consisting of an alkylating agent, an anthracycline, an epothilone, a histone deacetylase inhibitor, a Topoisomerase I inhibitor, a Topoisomerase II inhibitor, a kinase inhibitor, a nucleotide analog, nucleotide precursor analogs, a peptide antibiotic, a platinum-based anti-neoplastic, a retinoid, and a vinca alkaloid.
9. The method of claim 1 , wherein the treatment comprises the immunotherapy.
10. The method of claim 9 , wherein the immunotherapy is selected from the group consisting of a cellular therapy, an antibody therapy, and a cytokine therapy.
11. The method of claim 1 , further comprising calculating the prediction score λ, according to:
λ
=
1
K
[
log
L
(
θ
^
,
t
^
|
M
,
N
)
-
L
(
θ
=
0
|
M
,
N
)
]
wherein L is a maximum likelihood function, wherein {circumflex over (θ)} is an optimized value of the ctDNA burden coefficient θ, wherein {circumflex over (t)} is an optimized value of the cancer type t, wherein M is a function of X k , and wherein Nis a function of X k .
12. A method for treating a subject with a cancer, comprising:
determining a set of K DNA methylation marker regions of a cancer type t, K being a positive integer, wherein the set of K DNA methylation marker regions comprises CpG clusters that correspond to DNA methylation markers of the cancer type t;
obtaining a tumor class methylation pattern of each of the K DNA methylation marker regions, wherein the tumor class methylation pattern is expressed as m k T , wherein m denotes a marker region, wherein T denotes a tumor class, wherein k=1, 2, . . . K, and wherein the tumor class methylation pattern comprises a DNA methylation pattern derived from tumor cells of the cancer type t;
retrieving a normal class methylation pattern of each of the K DNA methylation marker regions, wherein the normal class methylation pattern is expressed as m k N , wherein m denotes a marker region, wherein N denotes a normal class, wherein k=1, 2, . . . K, and wherein the normal class methylation pattern comprises a cfDNA methylation pattern derived from cfDNA of non-cancerous cells;
obtaining a number of sequencing reads N of a cell-free DNA (cfDNA) methylation profile from a cfDNA sample of the subject, N being a positive integer;
calculating determining a circulating tumor DNA (ctDNA) burden θ based at least in part on the sequencing reads of the cfDNA methylation profile, the tumor class methylation pattern m k T , and the normal class methylation pattern m k N , wherein k=1, 2, . . . K;
determining that the subject has the cancer type t, when the ctDNA burden θ is greater than a pre-determined threshold; and
administering a treatment to the subject based on the determining that the subject has the cancer type t, wherein the treatment comprises a member selected from the group consisting of a chemotherapy, a radiation therapy, an immunotherapy, and a tumor resection.
13. The method of claim 12 , further comprising:
determining a methylation status in a sequencing read, the methylation status of the sequencing read being expressed as r=(r 1 , r 2 , . . . , r j , . . . , r L ), wherein j=1, 2, . . . L, wherein r j is binary, wherein r j =1 represents a CpG at site j being methylated in the sequencing read, wherein r j =0 represents a CpG at site j being unmethylated in the sequencing read, and wherein r j =NA represents a CpG at site j as not being covered by a sequencing read.
14. The method of claim 13 , wherein the methylation status r j of a CpG at site j in a sequencing read of the cfDNA methylation profile comprises a Bernoulli distribution expressed as r j ˜Bernoulli(p), wherein p is a methylation rate of a CpG at site j, wherein p follows a Beta prior distribution.
15. The method of claim 14 , further comprising:
calculating a likelihood P(r|m) of the methylation status of the sequencing read r=(r 1 , r 2 , . . . ) given a methylation pattern m, wherein the likelihood is calculated as P(r|m)=Π j P(r j |Beta(η,ρ))=Π j ∫ 0 1 Bernoulli(r j |p)Beta(p|η,ρ)dp, wherein Beta(η,η) is a beta function, and wherein p is an average methylation rate of a CpG at site j.
16. The method of claim 13 , further comprising calculating a likelihood P(r|m) of the methylation status of the sequencing read r=(r 1 , r 2 , . . . ) given a methylation pattern m, wherein the likelihood is calculated as
P
(
r
|
m
)
=
∏
i
∫
0
1
p
r
j
(
1
-
p
)
1
-
r
j
p
η
-
1
(
1
-
p
)
ρ
-
1
B
(
η
,
ρ
)
dp
=
∏
j
B
(
r
j
+
η
,
1
-
r
j
+
ρ
)
B
(
η
,
ρ
)
,
wherein Beta(η,ρ) is a beta function, and wherein p is an average methylation rate of a CpG at site j.
17. The method of claim 13 , wherein the methylation status r j in the number of sequencing reads N of the cfDNA methylation profile comprises a CpG site model that does not consider inter-individual variances.
18. The method of claim 13 , wherein the methylation status r j in the number of sequencing reads N of the cfDNA methylation profile comprises a bin model that does not consider inter-individual variances.
19. The method of claim 13 , wherein the methylation status r j in the number of sequencing reads N of the cfDNA methylation profile comprises a bin model that considers inter-individual variances.
20. The method of claim 12 , wherein the ctDNA burden θ is calculated determined using a maximum likelihood estimation P(R|θ, ),
wherein R denotes a set of N sequencing reads of the cfDNA methylation profile, expressed as R={r (1) , . . . , r (N) },
wherein denotes methylation patterns of the set of K DNA methylation marker regions in the cfDNA methylation profile, expressed as ={(m 1 T , m 1 N ), . . . , (m k T , m k N ), . . . , (m k T , m k N )}, and wherein P(R|θ, )=Π i=1 N (θP(r (i) |m k T )+(1−θ)P(r (i) |m k N )).
21. The method of claim 12 , wherein the treatment comprises the chemotherapy.
22. The method of claim 21 , wherein the chemotherapy is selected from the group consisting of an alkylating agent, an anthracycline, an epothilone, a histone deacetylase inhibitor, a Topoisomerase I inhibitor, a Topoisomerase II inhibitor, a kinase inhibitor, a nucleotide analog, nucleotide precursor analogs, a peptide antibiotic, a platinum-based anti-neoplastic, a retinoid, and a vinca alkaloid.
23. The method of claim 12 , wherein the treatment comprises the immunotherapy.
24. The method of claim 23 , wherein the immunotherapy is selected from the group consisting of a cellular therapy, an antibody therapy, and a cytokine therapy.
25. The method of claim 12 , wherein the normal class comprises a plurality of different non-cancerous tissue types.
26. The method of claim 12 , wherein the cfDNA of non-cancerous cells is obtained from a plurality of non-cancerous subjects.
27. The method of claim 12 , further comprising determining a biological composition of the cfDNA sample based at least partially on the number of sequencing reads N of the cfDNA methylation profile.
28. The method of claim 27 , wherein the biological composition of the cfDNA sample comprises a plurality of tissues.
29. The method of claim 28 , wherein the plurality of tissues is modeled by J known classes with methylation signatures Ω J (1≤j≤J) and an unknown tissues type Ω J+1 that models cfDNA unlikely to belong to known classes of tissue types.
30. The method of claim 29 , wherein determining the biological composition comprises solving a composition vector θ=(θ 1 , θ 2 , . . . , θ J , θ J+1 ,) where, wherein θ j (1≤j≤J) is the cfDNA proportion of tissue j in the cfDNA sample and Σ j=1 J+1 θj=1.
31. The method of claim 28 , wherein the plurality of tissues comprises at least two of: liver tissue, lung tissue, kidney tissue, colon tissue, T-cells, B-cells, neutrophils, small intestines tissue, pancreas tissue, adrenal glands tissue, esophagus tissue, adipose tissue, heart tissue, brain tissue, and placenta tissue.