Somatic mutation detection apparatus and method with reduced sequencing platform-specific error

Patent metadata

Field	Value
Publication number	US-11640662-B2
Application number	US-201917261164-A
Country	US
Kind code	B2
Filing date	Oct 25, 2019
Priority date	Oct 25, 2019
Publication date	May 2, 2023
Grant date	May 2, 2023

Abstract

A mutation detection apparatus includes a memory configured to store software for implementing a neural network and a processor configured to detect a mutation by executing the software, wherein the processor is configured to generate first genome data extracted from a target tissue and second genome data extracted from a normal tissue, extract image data by preprocessing the first genome data and the second genome data, and detect a mutation of the target tissue on the basis of the image data through the neural network trained to correct a sequencing platform-specific false positive.

First claim

The invention claimed is:

1. A mutation detection apparatus comprising:
a memory configured to store software for implementing a neural network; and
a processor configured to detect a mutation by executing the software,
wherein the processor is configured to:
generate first genome data extracted from a target tissue and second genome data extracted from a normal tissue;
extract image data by preprocessing the first genome data and the second genome data; and
detect a mutation of the target tissue on the basis of the image data through the neural network trained to correct a sequencing platform-specific false positive.

2. The mutation detection apparatus of claim 1 , wherein the neural network is trained to distinguish normal mutations from misdetected mutations on the basis of first training image data indicating training data on the normal mutations, which are normally detected, and second training image data indicating training data on the misdetected mutations, which are due to the false positive.

3. The mutation detection apparatus of claim 2 , wherein the first training image data and the second training image data are generated based on results of performing long-read sequencing and short-read sequencing on the same training tissue.

4. The mutation detection apparatus of claim 2 , wherein the first training image data and the second training image data include at least one of gene sequence, insertion and deletion (indel), and mapping quality.

5. The mutation detection apparatus of claim 1 , wherein the neural network is a convolutional neural network (CNN) configured to extract features from the image data and compute a probability that genes of the target tissue correspond to mutations on the basis of the features.

6. The mutation detection apparatus of claim 1 , wherein the processor performs preprocessing by correcting the first genome data and the second genome data on the basis of mapping quality and depth.

7. The mutation detection apparatus of claim 1 , wherein the mutation detected from the target tissue is a somatic single nucleotide variant (sSNV).

8. A method of detecting a mutation by executing software for implementing a neural network, the method comprising:
generating first genome data extracted from a target tissue and second genome data extracted from a normal tissue;
extracting image data by preprocessing the first genome data and the second genome data; and
detecting a mutation of the target tissue on the basis of the image data through the neural network trained to correct a sequencing platform-specific false positive.

9. The method of claim 8 , wherein the neural network is trained to distinguish normal mutations from misdetected mutations on the basis of first training image data indicating training data on the normal mutations, which are normally detected, and second training image data indicating training data on the misdetected mutations, which are due to the false positive.

10. The method of claim 9 , wherein the first training image data and the second training image data are generated based on results of performing long-read sequencing and short-read sequencing on the same training tissue.

11. The method of claim 9 , wherein the first training image data and the second training image data include at least one of gene sequence, insertion and deletion (indel), and mapping quality.

12. The method of claim 8 , wherein the neural network is a convolutional neural network (CNN) configured to extract features from the image data and compute a probability that genes of the target tissue correspond to mutations on the basis of the features.

13. The method of claim 8 , wherein the extracting of the image data comprises performing the preprocessing by correcting the first genome data and the second genome data on the basis of mapping quality and depth.

14. The method of claim 8 , wherein the mutation detected from the target tissue is a somatic single nucleotide variant (sSNV).

Assignees

• Seoul Nat Univ R&Db Foundation

Inventors

• Beak Dae Hyun
• Ahn Jun Hak
• Jeon Hyeon Seong
• Kim Do Yeon

Classifications

• G06N3/09

  Supervised learning · CPC title

• G06N3/0464

  Convolutional networks [CNN, ConvNet] · CPC title

• G06V2201/04

  Recognition of patterns in DNA microarrays · CPC title

• G06T2207/30072

  Microarray; Biochip, DNA array; Well plate · CPC title

• G06T2207/30024

  Cell structures in vitro; Tissue sections in vitro · CPC title