Chromosome CNV Detection with Fixed-Reference PCA Denoising
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for detecting chromosome copy-number variants (CNVs) in non-invasive prenatal DNA testing are inaccurate in determining the source of CNVs, particularly when the mother carries the variant, leading to false positives or negatives, and do not account for mosaicism and incomplete penetrance.
Innovation Solution
A method utilizing PCA denoising with a fixed reference set and hidden Markov models to analyze cell-free DNA sequencing data, combined with CNV prediction and filtration, to accurately determine the source of CNVs by calculating copy-number ratios and setting thresholds based on fetal concentration.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If traditional PCA denoising is applied to cell-free DNA sequencing data, then noise reduction is achieved, but false positive CNV detection occurs when the mother carries CNV
Solution Approach 1:
The patent applies PCA denoising as a preliminary step before CNV detection, but modifies it by using a fixed reference set from normal samples to establish noise patterns in advance. This preliminary action captures the characteristic that maternal CNV samples have different noise structures than normal samples, preventing the denoising process from incorrectly treating maternal CNV signals as noise.
Solution Approach 2:
The patent changes the parameter of the reference set from dynamic (sample-specific) to fixed (normal sample-based). By using a fixed reference set composed of normal samples, the PCA model establishes a stable baseline of normal noise patterns, allowing it to distinguish between actual noise and maternal CNV signals during the denoising process.
2Reliability
If current CNV detection methods are used, then CNV presence is detected, but the source of CNV cannot be accurately determined when the mother carries the variant
Solution Approach 1:
The patent segments the CNV detection process into distinct phases: (1) PCA denoising with fixed reference set, (2) HMM-based CNV detection, and (3) source determination through fetal concentration comparison. This segmentation allows each phase to optimize for its specific function, with the source determination phase specifically comparing detected CNVs against the known fetal concentration to distinguish fetal from maternal origins.
Solution Approach 2:
The patent introduces fetal concentration as an intermediary parameter to mediate between CNV detection and source determination. By comparing the detected CNV signals with the independently determined fetal concentration, the method can infer whether CNVs originate from the fetus or mother, resolving the ambiguity in source determination.
3Ease of operation
If non-invasive DNA detection is performed on cell-free DNA, then fetal CNV can be detected, but false positives occur when the sample carries large abnormal fragments such as abnormal maternal chromosome number
Solution Approach 1:
The patent converts the previously harmful effect (PCA mistakenly learning abnormal information from maternal CNV samples) into a beneficial feature by using a fixed reference set of normal samples. This approach causes PCA to consistently identify and remove normal noise patterns while preserving abnormal signals from maternal CNV, effectively using the consistency of normal samples to highlight abnormalities.
Solution Approach 2:
The patent changes the parameter of the PCA reference from sample-specific to fixed-normal, which fundamentally alters the denoising behavior. This parameter change ensures that the denoising process is consistent across all samples and does not adapt to abnormal patterns, thereby preserving maternal CNV signals while removing normal noise.
Data Source
Figure 1~2
Figure 3~4
Figure 5~6
AI summary
The present invention provides a method for detecting a chromosome copy-number variant (CNV). The method comprises: performing PCA noise reduction and CNV analysis on sequencing data of a sample to be detected, so as to detect whether said sample contains a CNV and/or determine a CNV source, wherein the PCA noise reduction is performed by comparing the sequencing data of said sample with a reference data set, and the reference data set is a principal component feature which represents noise and is obtained after PCA learning is performed on a predetermined sample. According to the method, a CNV can be accurately detected, and the occurrence source of the CNV can be determined, so that false positive or false negative results caused by a maternal CNV in a sample to be detected are effectively avoided, and false detection or missing detection of congenital defects of fetuses is prevented.