LOH Region Detection Model Using Low-Depth Genome Sequencing

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current methods for detecting Loss of Heterozygosity (LOH) regions are time-consuming and labor-intensive, requiring lengthy analysis steps and high-throughput technologies like Chromosome Microarray (CMA) that are costly and inefficient.

Innovation Solution

A method involving low-depth genome sequencing data derived from high-depth data, utilizing a self-attention allocation algorithm to train a LOH region detection model, determining information values and feature matrices to iteratively improve detection efficiency.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If high-depth genome sequencing data and complex high-throughput methods like CMA are used to detect LOH regions, then detection accuracy is improved, but detection efficiency deteriorates due to lengthy analysis steps and high time cost

Engineering Contradiction:
ImproveLOH region detection accuracyVSAvoiddetection efficiency
Core Design Contradiction:
Measurement precisionVSProductivity

Solution Approach 1:

The patent changes the sequencing depth parameter from high-depth to low-depth genome sequencing, fundamentally altering the input data characteristics. This parameter change enables the use of simplified analysis algorithms while maintaining acceptable detection accuracy, thereby resolving the contradiction between accuracy and efficiency

Inventive Principle:
Principle #35Parameter changes

Solution Approach 2:

The patent replaces complex mechanical/high-throughput detection systems (CMA technology with multiple analysis steps) with a computational model-based approach. The LOH region detection model substitutes the need for complex wet-lab procedures and manual analysis, achieving both accuracy and efficiency improvements

Inventive Principle:
Principle #28Mechanics substitution (Replace mechanical system)

2Measurement precision

If high-depth genome sequencing data is used for LOH region detection, then detection accuracy is improved, but time cost increases due to lengthy analysis steps

Engineering Contradiction:
ImproveLOH region detection accuracyVSAvoidanalysis time
Core Design Contradiction:
Measurement precisionVSLoss of time

Solution Approach 1:

The patent performs preliminary action by pre-training the LOH region detection model using high-depth sequencing data to learn accurate detection patterns. Once trained, the model can quickly process low-depth sequencing data without requiring lengthy analysis steps, thus reducing time loss while maintaining accuracy

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The patent creates a computational model that copies and encodes the detection knowledge from high-depth sequencing analysis. This model copy can then rapidly process new data without repeating the lengthy analysis steps of traditional methods, significantly reducing analysis time

Inventive Principle:
Principle #26Copying

3Measurement precision

If complex high-throughput methods are used for LOH region detection, then detection accuracy is improved, but labor cost increases

Engineering Contradiction:
ImproveLOH region detection accuracyVSAvoiddetection process simplicity
Core Design Contradiction:
Measurement precisionVSEase of manufacture

Solution Approach 1:

The patent implements self-service by enabling the LOH region detection model to automatically process sequencing data and identify LOH regions without requiring manual consultation of literature or databases. The model serves itself by making autonomous detection decisions, eliminating labor-intensive steps while maintaining accuracy

Inventive Principle:
Principle #25Self-service

Data Source

PatentEP4730338A1Training method and apparatus for loss of heterozygosity region detection model, device, and medium
Publication Date: 2026.04.22 TIANJIN MEDICAL LAB BGI
  • EP4730338A1 patent drawingFigure 1
  • EP4730338A1 patent drawingFigure 2
  • EP4730338A1 patent drawingFigure 3~4

AI summary

Embodiments of the present disclosure provide a method for training a loss of heterozygosity region detection model, an apparatus, a device and a medium, relating to the technical field of computers. The method includes: acquiring first data; determining information values of each region of N regions in a chromosome corresponding to the first data, wherein the information values include mean value information and variance information of the above regions; the N regions including heterozygous regions and the LOH regions, where N is a positive integer; determining feature information of the N regions, and determining an input feature matrix based on the feature information of the N regions and the information values of the N regions, wherein the feature information includes the information of the LOH regions corresponding to the N regions and an association relationship between the information values of each region; and performing iterative training on an initial model based on the input feature matrix to obtain a LOH region detection model, wherein the initial model is determined based on a self-attention allocation algorithm.