A resampling strategy for studying robustness in virus detection pipelines.

Area of Science:

• Bioinformatics
• Virology
• Computational Biology

Background:

• Next-generation sequencing (NGS) is crucial for identifying viral sequences in host samples.
• Current virus detection pipelines rely on mapping sequence reads to reference genomes, which can be uncertain due to sequence similarity and errors.
• This uncertainty affects the robustness of the identified viral lists.

Purpose of the Study:

• To introduce a novel resampling approach for assessing the robustness of virus detection results.
• To provide a method for evaluating the reliability of identified viral sequences and overall viral content analysis.
• To enhance the accuracy and confidence in viral identification from sequencing data.

Main Methods:

• Generation of artificial sequencing reads based on statistical distributions derived from original mapping results (SAM files).
• Resampling pipeline where artificial reads are mapped against reference genomes.
• Derivation of robustness indicators, including correlation of read counts and outlier detection, with visualization via Sankey diagrams.

Main Results:

• The resampling approach identified viruses that remained robustly detected, while others dropped from the original list.
• Demonstrated application on real-world and simulated data, including Influenza sequences.
• The method effectively highlights the stability and significance of detected viral signals.

Conclusions:

• The proposed resampling approach significantly improves the analysis of viral content in biological samples.
• It provides a reliable method for rating the robustness of initial findings in virus detection pipelines.
• The technique is adaptable to various read-mapping-based virus detection strategies.