Direct SNR Metric Calculation for Sequencing Base Calls
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing sequencing systems suffer from inaccuracies in determining signal-to-noise-ratio (SNR) metrics, leading to poor quality scores and inconsistent results due to delayed convergence and propagation of errors in early sequencing cycles, affecting the accuracy of nucleobase calls and other sequencing functions.
Innovation Solution
An improved method for determining per-cluster signal-to-noise-ratio metrics directly from maximum-likelihood model estimates during each sequencing cycle, using a direct approach to calculate noise levels without relying on rolling averages from previous cycles, allowing for accelerated convergence and improved accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If existing systems use rolling average of noise levels from previous cycles to determine SNR metrics, then the system can maintain continuity in SNR calculations, but the SNR metrics suffer from delayed convergence and error propagation in early sequencing cycles
Solution Approach 1:
The patent applies preliminary action by determining SNR metrics directly from maximum likelihood model estimates at each sequencing cycle without waiting for rolling averages from previous cycles. This allows the system to have accurate SNR metrics from the very first cycle rather than suffering from delayed convergence, as the necessary statistical parameters are computed upfront and used immediately for quality scoring
Solution Approach 2:
The patent extracts the SNR metric determination from the rolling average process and performs it independently using maximum likelihood estimates. By separating the SNR calculation from the iterative rolling average approach, the system eliminates error propagation while maintaining the benefits of statistical modeling, achieving both accuracy and fast convergence
2Measurement precision
If existing systems indirectly determine SNR metrics using multi-step processes with outside values, then the system can incorporate various intensity value attributes, but the process becomes complex and produces inaccurate quality scores
Solution Approach 1:
The patent merges the SNR metric determination into the maximum likelihood model estimation process itself. Instead of performing separate multi-step calculations with outside values, the system combines intensity value attribute analysis and SNR calculation into a unified statistical framework, reducing complexity while improving accuracy through consistent parameter estimation
Solution Approach 2:
The maximum likelihood model estimates serve as an intermediary that directly connects intensity value observations to SNR metrics. This intermediary approach allows the system to incorporate various intensity attributes without requiring complex multi-step processing, as the statistical model naturally integrates these factors into accurate SNR determinations
3Reliability
If existing systems treat early-cycle SNR metrics differently through weighting or omission, then the system attempts to accommodate delayed convergence, but the results become poor and inconsistent
Solution Approach 1:
The patent applies homogeneity by using the same maximum likelihood-based SNR determination method for all sequencing cycles without special treatment for early cycles. This uniform approach eliminates the need for complex weighting schemes or selective omission, producing consistent and reliable quality scores across the entire sequencing run while maintaining mathematical rigor
Data Source
AI summary
This disclosure describes methods, non-transitory computer readable media, and systems that generate an improved signal-to-noise-ratio metric for light signals emitted from fluorescent tags of nucleotide bases during a sequence run and use such signal-to-noise-ratio metrics to determine more accurate and flexible base calls. For instance, the disclosed systems can detect a series of signals from labeled nucleotide bases of a nucleotide-sample slide, determine intensity correction parameters based on intensity values for a given sequencing cycle, determine a scaling factor and a noise level based on the intensity correction parameters, and generate a signal-to-noise-ratio metric for the given sequencing cycle based on the scaling factor and the noise level. The disclosed systems can further utilize the signal-to-noise-ratio to generate quality metrics for generated nucleobase calls.