Measuring reproducibility of virus metagenomics analyses using bootstrap samples from FASTQ-files

Area of Science:

• Bioinformatics
• Genomics
• Virology

Background:

• High-throughput sequencing (HTS) data are susceptible to technical errors affecting experimental reproducibility.
• In virus metagenomics, these errors can lead to the misidentification of viruses, impacting study accuracy.
• Assessing the robustness of HTS data analysis is crucial for reliable scientific conclusions.

Purpose of the Study:

• To introduce a novel bootstrap resampling approach for generating artificial replicates of sequencing runs.
• To evaluate a mixture model for identifying potentially false positive findings based on virus read counts.
• To enhance the reliability and reproducibility of virus metagenomics analyses.

Main Methods:

• Developed a resampling approach using bootstrap sampling of sequencing reads from FASTQ files.
• Applied a mixture model to the distribution of read counts per virus.
• Evaluated the methods on both simulated and real-world virus metagenomics datasets.

Main Results:

• The bootstrap resampling approach demonstrated high reproducibility in virus detection, with strong correlation between original and bootstrap read counts.
• Bootstrap read counts effectively indicated variations in the evidence for virus presence.
• The mixture model showed good fit to read count distributions and improved accuracy in identifying true positives.

Conclusions:

• The proposed bootstrap resampling method and mixture model are effective tools for assessing the robustness of virus metagenomics analyses.
• These methods aid in identifying potential false positive findings and improving the reliability of virus detection.
• The RESEQ tool facilitates the practical application of bootstrap read generation for robust bioinformatics analysis.