Method and apparatus for training a machine learning model to detect true positive mutations in cell samples

Abstract

This disclosure relates to a method for training a machine learning model. The method for training a machine learning model includes the steps of: obtaining reference variant candidate information of a reference sample; generating annotation information associated with the reference variant candidate; generating training data based on the obtained reference variant candidate information and the generated annotation information; and training the machine learning model using the generated training data.

Claims

1. A method for training a machine learning model, which is executed by at least one processor, The steps include obtaining reference mutation candidate information for the reference sample, A step of generating annotation information associated with a reference mutation candidate, The steps include generating training data based on the acquired reference mutation candidate information and the generated annotation information, The steps include training a machine learning model using the generated training data and Methods for training machine learning models, including [specific examples].

2. The aforementioned reference sample is, Includes reference normal and reference abnormal samples taken from the same individual, The aforementioned acquired reference mutation candidate information is A method for training a machine learning model according to claim 1, which is determined based on first reference sequencing data associated with the reference normal sample and second reference sequencing data associated with the reference abnormal sample, using a mutation detection module.

3. The mutation detection module includes a plurality of detection modules, The aforementioned acquired reference mutation candidate information is This is determined by integrating (union) the reference mutation subcandidate information obtained using the aforementioned multiple detection modules. The acquired reference mutation sub-candidate information is as follows: A method for training the machine learning model according to claim 2, which is determined by applying the first reference sequencing data and the second reference sequencing data to each of the plurality of detection modules.

4. The step of generating the annotation information is: The steps include determining a plurality of reads in which at least a portion of the mapped locations overlap with the locations of the reference mutation candidate, The steps include generating first annotation information associated with the multiple reads determined above, and A method for training the machine learning model described in claim 1, including the following:

5. The aforementioned multiple reads include multiple variant reads that differ from the reference genome. The aforementioned first annotation information is, The minimum insert size of the plurality of mutant reads, the maximum insert size of the plurality of mutant reads, or the number of paired reads among the plurality of mutant reads that satisfy a specific condition, is included in at least one of these. The aforementioned specific conditions are, A method for training a machine learning model according to claim 4, comprising the conditions that the first and second leads of the paired reads are aligned in the forward and reverse directions, respectively, and the insert size of the paired reads is between a lower threshold and an upper threshold.

6. The step of generating the annotation information is: The steps include receiving normal tissue genome data (PON: Panel of Normals) generated from multiple sequencing data associated with multiple normal samples, The step of generating second annotation information linked to the aforementioned normal tissue genome data. A method for training the machine learning model described in claim 1, including the following:

7. The step of generating the annotation information is: The steps include receiving FFPE-treated tissue genome data (POF: Panel of FFPEs) generated from multiple sequencing data associated with multiple FFPE (Formalin-Fixed, Paraffin-Embedded)-treated samples, The step of generating third annotation information associated with the FFPE-treated tissue genome data. A method for training the machine learning model described in claim 1, including the following:

8. The aforementioned third annotation information is, A method for training a machine learning model according to claim 7, comprising the number of samples among the FFPE-treated samples in which the Variant Allele Frequency (VAF) for a specific position on the base sequence within the sample is less than a predetermined threshold.

9. The aforementioned third annotation information is, A method for training a machine learning model according to claim 7, comprising the number of samples among the FFPE-treated samples that have a predetermined number of mutant reads at specific positions on the nucleotide sequence within the sample.

10. The step of generating the annotation information is: The steps include generating fourth annotation information that includes information related to the variant type of the reference variant candidate and sequence context information of the reference variant candidate, and A method for training the machine learning model described in claim 1, including the following:

11. The step of generating the aforementioned training data is: A step of labeling classification information for the aforementioned reference mutation candidate. A method for training the machine learning model described in claim 1, including the following:

12. The aforementioned reference sample is an FFPE-treated sample. The labeling step described above is: The process includes the step of labeling the reference variant candidate as a true positive variant in response to the determination that at least a portion of the reference variant candidate information corresponds to at least a portion of the information associated with any one variant candidate among the FF (Fresh-Frozen) processed sample variant candidates. A method for training a machine learning model according to claim 11, wherein the FF-processed sample is a sample corresponding to the FFPE-processed sample.

13. The labeling step described above is: A method for training a machine learning model according to claim 12, further comprising the step of labeling the reference variant candidate as a false positive variant in response to the determination that the reference variant candidate information and the information associated with any one of the variant candidates in the FF-processed sample do not correspond to each other.

14. The step of generating the aforementioned training data is: A step of extracting the characteristics of the reference mutation candidate based on the reference mutation candidate information and the annotation information, The steps include including a dataset in the training data that includes the reference variant candidate information, the characteristics of the extracted reference variant candidate, and the labeled classification information. A method for training the machine learning model according to claim 11, further comprising:

15. The aforementioned machine learning model includes multiple classifiers. The step of training the aforementioned machine learning model is: The steps include inputting the reference mutation candidate information and the characteristics of the reference mutation candidate into each of the multiple classifiers, The steps include determining a classification result indicating whether the reference mutation candidate is a true positive mutation or not, using the output result from at least one of the plurality of classifiers, The steps include adjusting the parameters of the machine learning model based on the classification results and the classification information labeled to the reference variant candidate. A method for training the machine learning model according to claim 14, including the following:

16. The aforementioned machine learning model, A method for training a machine learning model according to claim 1, which receives target mutation candidate information and characteristics of the target mutation candidate within a target sample, and outputs a classification result indicating whether or not the target mutation candidate is a true positive mutation.

17. The aforementioned target sample is This includes both target normal and target abnormal samples collected from the same individual. The aforementioned target mutation candidate information is A method for training a machine learning model according to claim 16, which is determined based on first target sequencing data associated with the target normal sample and second target sequencing data associated with the target abnormal sample, using a mutation detection module.

18. The method for training a machine learning model according to claim 16, wherein the target sample is an FFPE-treated sample.

19. A genome profiling method performed by at least one processor, through detection of true positive mutations within a cell sample, The steps include obtaining target mutation candidate information for the target sample, The steps include: using a machine learning model to determine a classification result indicating whether a target mutation candidate is a true positive mutation or not; The steps include: performing genomic profiling on the target sample based on the determined classification results; Includes, The aforementioned machine learning model, A genome profiling method trained to determine whether a candidate reference mutation is a true positive mutation, using reference mutation candidate information from a reference sample and annotation information associated with the reference mutation candidate.

20. The genome profiling method according to claim 19, further comprising the step of providing at least one of disease diagnostic information, treatment strategy information, prognosis prediction information, or drug response prediction information of the individual from whom the target sample was collected, based on the results of performing the genome profiling.

21. A computer-readable non-temporary recording medium that stores instruction words for executing the method according to claim 1 on a computer.

22. It is a device, Memory and Connected to the memory and configured to execute at least one computer-readable program contained in the memory, Includes, The aforementioned at least one program, Obtain reference mutation candidate information for the reference sample, Annotation information linked to the aforementioned reference mutation candidate information is generated, A device comprising instructions for generating training data based on the acquired reference variant candidate information and the generated annotation information, and for training a machine learning model using the generated training data.

Images