High-resolution melt curve classification using neural networks
The present invention relates to a method and system for classifying high-resolution melt (“HRM”) curves, and, more specifically, to a method and system for classifying HRM curves by genotype where the curves are represented by a mathematical function with varying coefficient values.
1. A method for classifying high resolution melt (“HRM”) curve data by genotype, the method comprising:
obtaining melt curve data from double-stranded DNA from an obtained biological sample of interest by heating the obtained biological sample of interest and measuring a resulting change in fluorescence as a function of a temperature of the obtained biological sample of interest;
fitting, by a processor, a corresponding mathematical function to at least one melt curve represented by melt curve data, wherein the step of fitting fits the data with a Chebyshev polynomial expansion;
determining, by the processor, the mathematical function's coefficient values;
inputting the coefficient values into a classification tool for classification comprising a neural network, wherein the neural network comprises at least one of the following neuron types:
input neurons, where the number of neurons equals the number of coefficient values;
one or more hidden layers of neurons; and
plurality of output neurons, wherein each of said output neurons represents a class comprising a genotype or a group of indistinguishable genotypes that unknown HRM curve data will be assigned to; and
classifying the melt curve data by the neural network as either a known genotype or an unknown genotype based on the inputted coefficient values.
2. The method of claim 1 , wherein the step of fitting, by the processor, the mathematical function's coefficient values further comprises determining the mathematical function's Chebyshev coefficient values.
3. The method of claim 2 , wherein the step of using a Chebyshev polynomial expansion is pursuant to the following equation:
T 0 ( x )=1
T 1 ( x )= x
T n+1 ( x )=2 ×T n ( x )− T n−1 ( x )
where T0(x) is the Chebyshev polynomial at order 0, and T1(x) is the Chebyshev polynomial at order 1.
4. The method of claim 1 , wherein the step of using the coefficient values as an input for classification in a classification tool further comprises the step of constructing a training data set prior to designing and implementing the classification tool, wherein the training data set comprises HRM curve data representing a population of genotypes to be classified.
5. The method of claim 1 , further comprising the step of interpreting output values of the neural network as the probability that an unknown HRM curve data belongs to a given class, if the coefficient values in the training set are normally distributed.