Method and system for analysis of melt curves, particularly dsDNA and protein melt curves

1. A computer-implemented method of analysis of melt curve data, comprising the steps of:

(a) in a computer memory, parameter model fitting the melt curve data;

(b) in a computer memory, performing a principal component analysis of the melt curve data; and

(c) utilizing the principal components for clustering the melt curve data into groups.

2. The method of claim 1 , wherein the melt curve data is double stranded deoxyribonucleic acid (dsDNA) melt curve data for identifying a specific sequence of dsDNA in a sample after performing a polymerase chain reaction (PCR).

3. The method of claim 1 , wherein the computer memory is in a computer connected to the Internet or a network.

4. The method of claim 1 , wherein the computer comprises a storage medium and a monitor or visual display device.

5. The method of claim 1 , wherein the step (a) comprises fitting the melt curve data to a model of the form:

A

+

B

-

CT

1

+

exp

⁡

(

λ

⁡

(

T

-

T

0

)

)

where A is an estimate of the minimum observed fluorescence; B and C are estimates of the intercept and slope of the initial part of the melt curve; T 0 is estimated as the point of maximum (negative) slope of the melt curve, and γ is approximately 2.

6. The method of claim 1 , wherein the step (a) comprises a scaling of the melt curve to a line of best fit.

7. The method of claim 6 , wherein a region of melt curve data prior to the melt curve transition and a region of melt curve data following the melt curve transition are selected and an average fluorescence value is calculated.

8. The method of claim 1 , wherein at least the first two principal components are utilized in step(c).

9. The method of claim 1 , wherein said clustering occurs by means of k-means clustering or hierarchical clustering.

10. The method of claim 9 , wherein a gap statistic is utilized for determining the number of groups for the melt curve data.

11. The method of claim 10 , wherein determining the number of groups for the melt curve data comprises the comparison of a measure of cluster quality to data known not to have any real clusters.

12. The method of claim 11 , wherein the measure of cluster quality is determined applying the formula:

W

k

=

∑

j

=

1

k

⁢

∑

i

∈

c

j

⁢



x

i

-

x

_

j



2

where C j is the j-th cluster, and x j is its center.

13. The method of claim 12 , wherein determining the number of groups for the melt curve data comprises the steps of:

(i) clustering the observed data, varying the number of clusters k=1, 2, . . . , K for a maximum K and computing W k for each k;

(ii) generating B random reference data sets and clustering each as in step (i), computing W kb for k=1,2, . . . , K and b=1,2, . . . ,B;

(iii) computing the Gap statistic and standard error for each k applying the formulae

Gap

⁡

(

k

)

=

(

1

/

B

)

⁢

∑

log

⁡

(

W

k

.

⁢

⁢

b

)

-

log

⁡

(

W

k

)

⁢

⁢

and

s

k

=

(

1

+

1

/

B

)

⁢

{

(

1

/

B

)

⁢

∑

b

⁢

(

log

⁡

(

W

k

.

⁢

⁢

b

)

-

(

1

/

B

)

⁢

∑

b

⁢

log

⁡

(

W

k

.

⁢

⁢

b

)

)

2

;

⁢



⁢

and

(iv) choosing the number of clusters k to be the smallest {circumflex over (k)} such that Gap(k)≧Gap(k+1)−s k+1 .

14. The method of claim 1 , wherein a Bayes Information Criterion analysis is utilized for determining the number of groups for the melt curve data.

15. The method of claim 1 , wherein said melt curve data is for a known substance and said clustering utilises linear discriminant analysis to cluster the melt curve data into groups.

16. The method of claim 1 , wherein said clustering utilises a typicality index to assign the melt curve data into a group.

17. The method of claim 1 , wherein said melt curve and said clustering utilize posterior class probabilities to cluster the melt curve data into groups.

18. A computer-implemented system for analysis of melt curve data, comprising a non-transitory computer readable medium comprising:

import means to import raw high resolution melting data;

parsing means to syntactically analyze the imported data: and

cropping means to allow a user to avoid regions of non-specific amplification products.

19. The system of claim 18 , further comprising sample selection means to allow a user to select a sample, and mode selection means to allow a user to select between analysing a supervised data set or an unsupervised data set following the sample selection.

20. The system of claim 19 , further comprising parameter selection means to select a most appropriate cluster number and a most appropriate principal component number via k-means clustering, Gap statistics, or both, when the user selects analysing an unsupervised data set.

21. The system of claim 20 , further comprising control selection means to allow the user to select the appropriate controls for each genotype when the user selects analysing a supervised data set, wherein linear discriminant analysis is used and the appropriate number of principal components is determined via a cross-validation function if more than two controls are used for each group and wherein nearest neighbor classification of samples into groups which is proceeded by a re-computation using linear discriminant analysis is used if one single control activates.

22. A method of analysis of melt curve data, comprising the steps of:

(a) generating melt curve data with a thermal cycler;

(b) storing the melt curve data in an electronic file on a computer;

(a) in a computer memory, parameter model fitting the melt curve data;

(b) in a computer memory, performing a principal component analysis of the melt curve data; and

(c) utilizing the principal components for clustering the melt curve data into groups.