Base Coverage Normalization for Base-Level CNV Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current CNV detection algorithms in next-generation sequencing data are limited by regional coverage variations, making it difficult to accurately identify and correct biases below the region level, which hinders the detection of copy number variations in genomic regions.
Innovation Solution
A method that normalizes base level coverage by modeling expected coverage biases using training data from healthy individuals, correcting for biases at the individual nucleotide base level, and employing models to detect deviations indicative of CNVs, such as GC content, mappability, and principal component analysis.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If regional coverage normalization is used, then the detection of CNVs at region level is improved, but the ability to identify and correct sources of coverage variation below the region level deteriorates
Solution Approach 1:
The patent segments the coverage analysis from region level to base level. Instead of normalizing coverage at the regional level where variations are averaged out, the method performs normalization at the individual base level, allowing identification and correction of coverage biases at the finest granularity. This segmentation enables detection of coverage variations that were previously hidden at the region level.
Solution Approach 2:
The patent applies local quality by treating each base position with its own specific normalization factors rather than applying a uniform regional normalization. Coverage biases are modeled and corrected locally at each base position using base-specific factors derived from training data, allowing precise correction of local coverage variations while maintaining overall detection accuracy.
2Manufacturing precision
If base level coverage normalization is implemented, then the correction of coverage biases is improved, but the computational complexity increases
Solution Approach 1:
The patent applies preliminary action by pre-computing base-specific normalization factors using training data from healthy individuals before analyzing test samples. Coverage biases are modeled in advance during a training phase, and these pre-computed factors are then applied during test sample analysis. This separates the complex modeling work from the actual detection process, reducing real-time computational complexity while maintaining high correction accuracy.
Solution Approach 2:
The patent uses copying by creating a reference model of coverage biases from training data that represents the expected coverage pattern. This reference model is copied and applied to multiple test samples, avoiding repeated computation of normalization factors for each sample. The base-specific normalization factors are computed once and reused, significantly reducing computational complexity for batch processing.
3Measurement precision
If training data from healthy individuals is used for modeling, then the baseline coverage bias modeling is improved, but the requirement for additional data increases
Solution Approach 1:
The patent applies universality by using training data from healthy individuals to create a universal model of coverage biases that applies across different populations and applications. The base-specific normalization factors derived from healthy individuals serve multiple purposes: correcting technical biases in various test samples, enabling detection of CNVs in different disease contexts, and providing a reference for comparing different sequencing experiments. This single training set supports multiple detection scenarios.
Data Source
Figure 1A
Figure 1B
Figure 1C
AI summary
Gene copy number variations are identified for genes in a targeted gene panel. For each gene, coverage at each base position across the gene is determined. The coverage at each base position can be influenced by the hybridization probes that are used to determine the base level coverage of the base position. The base level coverage for each base position is normalized to account for the characteristics of the hybridization probes. To determine whether a copy number variation exists for a gene, the base level coverage of base positions across the gene for a subject is analyzed to determine whether it deviates from the base level coverage of base positions across the gene for previously analyzed, healthy individuals. If a significant deviation exists, a copy number variation for the gene is called.