Dynamic, adaptive sampling during nanopore sequencing using Bayesian experimental design.

Area of Science:

• Genomics
• Bioinformatics
• Molecular Biology

Background:

• Nanopore sequencing offers molecule selection capabilities.
• Current selection strategies are static, limiting experimental adaptability.
• Re-focusing sequencing efforts on high-value molecules is not possible with existing methods.

Purpose of the Study:

• Introduce BOSS-RUNS, an algorithmic framework for dynamic decision strategies in nanopore sequencing.
• Optimize information gain by dynamically updating molecule selection based on real-time data.
• Mitigate coverage bias and improve variant calling in complex samples.

Main Methods:

• Developed BOSS-RUNS, software for real-time uncertainty quantification at genome positions.
• Implemented a decision strategy to select DNA fragments based on expected uncertainty reduction.
• Applied BOSS-RUNS to microbial communities to assess its impact on coverage and variant detection.

Main Results:

• BOSS-RUNS dynamically updates molecule selection strategies based on observed data.
• Significantly reduced coverage bias in microbial communities, including low-abundance species.
• Achieved an 87.5% reduction in low-coverage sites and detected 12.5% more single-nucleotide polymorphisms.

Conclusions:

• BOSS-RUNS enables data-driven, adaptive molecule selection in nanopore sequencing.
• Improves sequencing efficiency and data quality, particularly for low-coverage or divergent regions.
• Applicable to diverse sequencing scenarios, enhancing variant calling and reducing time-to-answer.