QQ-SNV: single nucleotide variant detection at low frequency by comparing the quality quantiles

Area of Science:

• Genomics
• Bioinformatics
• Virology

Background:

• Next-generation sequencing (NGS) enables the study of heterogeneous viral infections.
• Deep sequencing allows for the detection of low-frequency variants.
• Distinguishing true variants from sequencing errors is crucial for accurate analysis.

Purpose of the Study:

• To develop and evaluate QQ-SNV, a novel logistic regression classifier for single nucleotide variant (SNV) calling.
• To improve the accuracy of SNV detection in viral deep sequencing data from Illumina platforms.
• To compare QQ-SNV's performance against existing methods like LoFreq, ShoRAH, and V-Phaser 2.

Main Methods:

• Developed QQ-SNV, a logistic regression model utilizing quality score quantiles to differentiate true SNVs from sequencing errors.
• Trained the model on an in silico mixture dataset of five HIV-1 plasmids.
• Tested QQ-SNV against established methods on HIV, HCV plasmid mixtures, and an influenza H1N1 clinical dataset.

Main Results:

• QQ-SNV(HS-P80) demonstrated superior accuracy by balancing sensitivity and specificity across all test sets.
• In an HCV sequencing study, QQ-SNV(HS-P80) achieved 100% sensitivity and 100% specificity for variants down to 0.5% frequency.
• QQ-SNV required the least computation time and consistently detected low-frequency variants in a clinical sample.

Conclusions:

• QQ-SNV is a novel and highly efficient method for single nucleotide variant calling in Illumina deep sequencing virology data.
• The method successfully balances sensitivity and specificity, outperforming existing tools.
• QQ-SNV offers a valuable advancement for analyzing complex viral populations.