Deletion Detection in Oligonucleotide Sequences via ML Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The detection of deletions in genomes using short reads is computationally expensive and inaccurate due to the need for alignment, which is challenging with existing sequencing technologies.
Innovation Solution
A method that splits each read into segments (k-mers) and trains a machine learning model directly on these segments without alignment, allowing for the detection of missing k-mers indicative of deletions, using a neural network with a bidirectional gated recurrent unit for accurate diagnosis.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If short reads are aligned against a reference genome to detect deletions, then deletion detection can be performed, but the computational complexity becomes excessively high and accuracy decreases when deleted regions are longer than a single read
Solution Approach 1:
The patent divides each sequencing read into multiple overlapping segments or sub-sequences. This segmentation allows the machine learning model to process smaller units individually, making the overall system more manageable and accurate in detecting deletions without requiring computationally expensive alignment of entire reads against the reference genome.
Solution Approach 2:
The patent replaces the traditional mechanical alignment process with a machine learning-based approach. Instead of aligning reads against a reference genome using computational alignment algorithms, the system uses a trained neural network model that directly processes segmented reads to predict deletions, significantly reducing computational complexity while maintaining or improving accuracy.
2Reliability
If alignment and variant calling are performed to detect deletions, then deletion detection is possible, but the process becomes computationally expensive and time-consuming
Solution Approach 1:
The patent performs preliminary training of the machine learning model using labeled training data before actual deletion detection. This preliminary action creates a pre-trained model that can quickly process new sequencing data without requiring time-consuming alignment and variant calling steps during the actual detection phase, thus reducing processing time while maintaining reliability.
Solution Approach 2:
The patent substitutes the traditional alignment and variant calling pipeline with a direct machine learning prediction approach. The trained model processes segmented reads and directly outputs deletion predictions, eliminating the need for computationally expensive and time-consuming alignment operations while preserving detection capability.
Data Source
AI summary
Disclosed herein is a method for detecting deletion in a gene sequence. The method comprises receiving, by a processor, training sequencing data, which comprises multiple training reads associated with gene sequences with deletion and gene sequences without deletion. The processor splits each of the multiple training reads into multiple training segments shorter than the training reads and trains a machine learning model on the multiple segments. The processor receives testing sequencing data comprising multiple testing reads, splits each of the multiple testing reads into multiple testing segments, and evaluates the trained machine learning model to the multiple testing segments to detect deletion in the testing sequencing data. No alignment or variant calling is necessary, which reduces the computational complexity of the evaluation step significantly.