Assessing the consequences of denoising marker-based metagenomic data

Area of Science:

• Microbial ecology
• Bioinformatics
• Genomics

Background:

• Early metagenomic studies often ignored noise from sequencing and PCR errors.
• Denoising algorithms were primarily validated on mock communities, not real-world data.

Purpose of the Study:

• To evaluate the impact of five denoising pipelines on a real 16S rRNA metagenomic dataset.
• To understand how these algorithms alter sequence reads during the denoising process.

Main Methods:

• Processed a real 16S rRNA metagenomic dataset through five different denoising pipelines.
• Reconstituted sequence reads at each stage to track alterations.
• Examined specific algorithms within pipelines like AmpliconNoise and QIIME.

Main Results:

• AmpliconNoise altered reads inconsistently with known pyrosequencing errors until parameters were adjusted.
• QIIME introduced numerous read changes unless default parameters were modified.
• Mothur's pipeline had fewer side-effects due to strict filtering but reduced sequence output.

Conclusions:

• Denoising pipelines can significantly alter metagenomic reads in ways not always intended.
• Users should carefully examine how denoising processes transform sequence data in real community studies.
• Awareness of these transformations is crucial for accurate metagenomic analysis.