Variant Call Recalibration for Multiallelic and Haploid Genomes

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing nucleotide base sequencing platforms and sequencing data analysis software inaccurately determine nucleotide base calls, particularly for multiallelic and haploid genomic coordinates, and often produce false negative variant calls, while being computationally inefficient and lacking interpretability.

Innovation Solution

A call recalibration machine learning model is employed to update sequencing metrics, generating variant call classifications that improve accuracy and efficiency by adapting to specific genomic coordinates, such as multiallelic, haploid, and falsely identified homozygous reference genotypes, using a combination of internal and external sequencing metrics.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If existing variant callers are used to determine nucleotide base calls, then variant calls can be generated, but accuracy is poor particularly for multiallelic and haploid genomic coordinates

Engineering Contradiction:
Improvenucleotide base call accuracyVSAvoidvariant call reliability
Core Design Contradiction:
Measurement precisionVSReliability

Solution Approach 1:

The patent introduces an intermediary processing step between raw sequencing data and final variant calls. A machine learning model acts as a mediator to recalibrate base calls by learning from both biallelic and multiallelic training data, then applying this knowledge to improve accuracy for challenging genomic coordinates including haploid and multiallelic regions.

Inventive Principle:
Principle #24Intermediary (Mediator)

Solution Approach 2:

The patent changes the parameters used for variant calling by training machine learning models with expanded datasets that include multiallelic coordinates. The model learns to adjust prediction parameters based on sequencing metrics and contextual information, improving accuracy for previously challenging genomic regions.

Inventive Principle:
Principle #35Parameter changes

2Adaptability or versatility

If machine learning based sequencing systems are used, then nucleotide base calls can be determined, but training data is largely biallelic leading to poor performance on multiallelic coordinates

Engineering Contradiction:
Improvegenomic coordinate type coverageVSAvoidmultiallelic coordinate accuracy
Core Design Contradiction:
Adaptability or versatilityVSMeasurement precision

Solution Approach 1:

The patent makes the machine learning model universal by training it on multiple types of genomic coordinates simultaneously. The model is designed to handle both biallelic and multiallelic coordinates, as well as haploid and diploid regions, using a unified architecture that adapts to different coordinate types through learned patterns from diverse training data.

Inventive Principle:
Principle #6Universality (Multi-functionality)

Solution Approach 2:

The patent performs preliminary action by pre-training the machine learning model on comprehensive datasets that include multiallelic coordinates before deployment. This pre-training ensures the model has learned appropriate patterns for handling diverse genomic coordinate types, improving its performance on multiallelic regions from the outset.

Inventive Principle:
Principle #10Preliminary action

3Reliability

If existing sequencing systems are used, then variant calls can be generated, but excessive false negative variant calls are identified

Engineering Contradiction:
Improvevariant call accuracyVSAvoidfalse negative rate
Core Design Contradiction:
ReliabilityVSProductivity

Solution Approach 1:

The patent implements feedback mechanisms where the machine learning model continuously learns from sequencing metrics and recalibrates base calls based on predicted classifications. The system uses feedback from training data and performance metrics to improve its ability to identify true variants while reducing false negatives in production sequencing.

Inventive Principle:
Principle #23Feedback

4Productivity

If conventional sequencing data analysis software is used, then nucleotide base calls can be determined, but computational efficiency is low and interpretability is lacking

Engineering Contradiction:
Improveprocessing speedVSAvoidmodel complexity
Core Design Contradiction:
ProductivityVSDevice complexity

Solution Approach 1:

The patent segments the variant calling process into distinct stages: initial base call determination, sequencing metric extraction, machine learning classification, and final variant call generation. This segmentation allows each component to be optimized independently, improving overall computational efficiency while maintaining interpretability through clear process boundaries.

Inventive Principle:
Principle #1Segmentation

Data Source

PatentEP4457822B1Machine learning model for recalibrating nucleotide base calls corresponding to target variants
Publication Date: 2025.12.17 ILLUMINA INC
  • EP4457822B1 patent drawingFigure 1
  • EP4457822B1 patent drawingFigure 2
  • EP4457822B1 patent drawingFigure 3A

AI summary

This disclosure describes methods, non-transitory computer readable media, and systems that can utilize a machine learning model to recalibrate nucleotide base calls (e.g., variant calls) of a call generation model. For instance, the disclosed systems can train and utilize a call recalibration machine learning model to generate a set of predicted variant call classifications based on sequencing metrics associated with a sample nucleotide sequence. Leveraging the set of variant call classifications, the disclosed systems can further update or modify nucleotide base calls (e.g., variant calls) corresponding to genomic coordinates, such as multiallelic genomic coordinates, haploid genomic coordinates, and genomic coordinates indicated (by the call generation model) to exhibit homozygous reference genotypes.