AI Base Calling for Reordered NGS Index Sequences
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Index misassignment between multiplexed libraries in next-generation sequencing (NGS) technologies complicates data analysis, as existing methods struggle with increased throughput and accuracy in base calling of index sequences.
Innovation Solution
Employing a neural network-based base caller that utilizes reordering of sequencing images and a specialized architecture to segregate processing of data for different sequencing cycles, combined with training on reordered image sets to enhance accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If existing base calling methods are used for multiplexed libraries, then processing speed is maintained, but base calling accuracy and handling of index sequence complexity deteriorate
Solution Approach 1:
The patent segments the base calling process by implementing separate processing pipelines for different sequencing cycles and index sequences. The neural network architecture processes each index sequence independently through dedicated layers, allowing accurate handling of complex multiplexed libraries while maintaining efficient parallel processing throughput.
Solution Approach 2:
The patent transforms the input data by reordering sequencing images and adjusting processing parameters to optimize both accuracy and throughput. The neural network dynamically adjusts its processing based on the complexity of index sequences, changing operational parameters to handle varying levels of multiplexing efficiency.
2Reliability
If traditional demultiplexing methods are used, then computational complexity is reduced, but index misassignment between multiplexed libraries increases
Solution Approach 1:
The patent replaces traditional mechanical demultiplexing algorithms with an artificial intelligence-based neural network system. This substitution enables the system to accurately resolve index misassignment issues in multiplexed libraries by learning complex patterns in sequencing data, achieving high reliability despite increased computational requirements.
Solution Approach 2:
The neural network acts as an intermediary between raw sequencing data and final demultiplexing results. It processes index sequences through multiple layers of computation, including specialized handling of index reads and target reads, to produce accurate library assignments while managing computational complexity through optimized architecture.
3Measurement precision
If neural network processing is applied to all sequencing data, then base calling accuracy improves, but processing time and computational resources increase
Solution Approach 1:
The patent applies neural network processing selectively rather than uniformly to all sequencing data. The system identifies and applies intensive AI-based base calling primarily to index sequences and complex regions, while using more efficient methods for simpler data portions, thereby improving accuracy where needed without proportionally increasing overall processing time.
Solution Approach 2:
The patent performs preliminary processing and filtering of sequencing data before applying the full neural network model. By pre-processing images and identifying key features in advance, the system reduces the computational burden on the neural network, maintaining high accuracy while minimizing additional processing time.
Data Source
AI summary
Techniques for improving artificial intelligence-based base calling are disclosed. The improved techniques can be used to better train artificial intelligence for base calling by reordering of sequencing images, and training of a neural network-based base caller where the temporal logic is effectively “frozen” (or bypassed). In addition, the improved techniques include various combinations, including, for example, combining “normalization” of sequencing images with reordering of sequencing images and/or with effectively “freezing” the temporal logic.