Methylation State Vectors for Duplicate Marking in cfDNA
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for analyzing DNA methylation sequencing data from cell-free DNA face challenges in accurately identifying unique fragments and distinguishing them from duplicates, leading to false positives and reduced accuracy in disease classification, particularly in cancer detection.
Innovation Solution
An analytics system generates methylation state vectors for DNA fragments, computes difference scores based on methylation states at CpG sites, and uses a modified Hamming distance to differentiate between unique and duplicate fragments, while applying a cancer prediction model to identify anomalous fragments indicative of cancer.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If traditional duplicate marking methods are used to identify duplicate DNA fragments, then duplicate removal efficiency is improved, but false positive rate increases and measurement precision deteriorates
Solution Approach 1:
The patent changes the parameters used for duplicate identification from simple sequence identity to include methylation state vectors. By incorporating methylation states at CpG sites as additional parameters, the system can distinguish between true duplicates and unique fragments with similar sequences, thereby reducing false positives while maintaining duplicate removal efficiency
Solution Approach 2:
The patent introduces methylation state vectors as an intermediary layer between sequence data and duplicate identification. These vectors capture epigenetic information that serves as a mediator to differentiate fragments, allowing the system to maintain high duplicate removal efficiency while reducing false positives through the additional discriminatory power of methylation patterns
2Measurement precision
If methylation state vectors are computed for all CpG sites to improve fragment differentiation, then measurement precision is improved, but computational complexity and device complexity increase
Solution Approach 1:
The patent segments the genome into specific CpG sites and processes methylation data only at these relevant locations rather than across the entire genome. This segmentation approach maintains high measurement precision for fragment differentiation while reducing computational complexity by focusing resources on biologically significant regions
Solution Approach 2:
The patent applies local quality by computing methylation state vectors specifically at CpG sites where methylation patterns are most informative for differentiation. Rather than uniformly processing all genomic positions, the system concentrates computational effort on local regions (CpG sites) that provide the highest discriminatory value, thereby balancing precision improvement with computational feasibility
3Reliability
If a modified Hamming distance with difference scores is used to differentiate fragments, then reliability of cancer detection is improved, but computational time and complexity increase
Solution Approach 1:
The patent performs preliminary computation of methylation state vectors and difference scores before the actual duplicate marking and cancer detection processes. By pre-calculating these values and storing them, the system reduces the computational time required during the main analysis workflow, thereby improving reliability without excessive time loss
Solution Approach 2:
The patent implements a threshold-based approach where only fragments with Hamming distances below a certain threshold require full methylation state comparison. This partial action strategy allows the system to quickly dismiss obviously different fragments while applying the more computationally intensive modified Hamming distance calculation only when necessary, thus improving cancer detection reliability while controlling computational time
Data Source
AI summary
An analytics system marks duplicate fragments from an initial set of fragments from a subject. The analytics system generates a sample state vector for each fragment. Each sample state vector comprises a sample genomic location within a reference genome and a plurality of methylation states for a plurality of CpG sites in the fragment, the methylation states determined to be one of methylated, unmethylated, variant, and ambiguous. The analytics system identifies two fragments with methylation state vectors as being derived from a matching reference location, e.g., sharing a common plurality of CpG sites. The analytics system calculates a modified Hamming distance based on methylation states in the first sample state vector and methylation states in the second sample state vector. Based on the modified Hamming distance, the analytics system marks the first fragment and the second fragment as either duplicate fragments or non-duplicate fragments.