Adaptive Base Calling for Sequencer Drift Correction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Nucleic acid sequencers face accuracy issues due to instrument drift over time, affecting the interpretation of signal intensity and homopolymer length calls in flow sequencing methods, leading to a need for efficient recalibration to maintain sequencing data quality.
Innovation Solution
The method involves updating a pre-trained sequencer-specific machine-learning model using sequencing data from previous runs, selecting a subset of data, calling preliminary sequences, mapping them to a reference sequence, and iteratively updating the model until a quality control threshold is met, allowing for efficient recalibration and improved processing speed.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If the sequencer operates continuously without recalibration, then productivity is maintained, but measurement precision deteriorates due to instrument drift
Solution Approach 1:
The system implements periodic recalibration of the machine learning model at predetermined intervals during sequencing operations. This allows the system to maintain measurement precision by periodically updating the model with new sequencing data, while minimizing interruptions to overall productivity since recalibration occurs during routine operational cycles rather than requiring complete system shutdowns.
Solution Approach 2:
The machine learning model performs self-updating by automatically recalibrating using sequencing data generated during normal operations. The system uses its own operational data to retrain and update the model, eliminating the need for external manual recalibration processes and maintaining continuous operation without requiring additional resources or personnel intervention.
2Measurement precision
If the machine learning model is updated frequently to maintain accuracy, then measurement precision is improved, but processing time increases
Solution Approach 1:
Instead of performing complete model retraining with all available data at each update cycle, the system applies partial updates using only the most recent sequencing data batches. This incremental learning approach maintains measurement precision by continuously adapting the model while significantly reducing the computational time and resources required compared to full retraining cycles.
Solution Approach 2:
The system prepares and pre-processes sequencing data in advance for model updates, organizing data batches and preparing training sets before they are needed for recalibration. This preliminary data preparation reduces the actual processing time during model updates by having clean, organized data ready for immediate use in the recalibration process.
3Measurement precision
If complete sequencing data is used for model updates, then measurement precision is improved, but device complexity increases
Solution Approach 1:
The system extracts and uses only the necessary portions of sequencing data for model updates, specifically selecting data batches that are most relevant for recalibration rather than processing complete datasets. This extraction of essential data elements maintains model calibration accuracy while reducing computational complexity and resource requirements by eliminating unnecessary data processing steps.
Solution Approach 2:
The large sequencing dataset is divided into smaller, manageable batches for model updates. Instead of processing complete datasets at once, the system segments data into incremental batches that can be processed efficiently, reducing memory requirements and computational complexity while maintaining the statistical power needed for accurate model recalibration.
Data Source
AI summary
Methods for updating a system comprising a sequencer are described herein. In some exemplary methods, the system is updated through generating sequencing data for a plurality of nucleic acid molecule colonies, selecting sequencing data for a subset of the nucleic acid molecule colonies, calling preliminary sequences for the subset of the nucleic acid colonies, mapping the called preliminary sequences to a known reference sequence, and updating the pre-trained sequencer-specific machine-learning model. Also described herein are systems for carrying out such methods and computer readable memory for storing such methods.