A-B.05: Impact of bioinformatic pipeline selection on microbial community inference: a comparative analysis across pipelines including MetagenApp
Author
Keywords
16S rRNA metabarcoding, microbiome bioinformatics, amplicon sequencing pipelines, MetagenApp, pipeline benchmarking, reproducibility
[sponser-meet-now-chat][/sponser-meet-now-chat]
Bioinformatic pipeline choice can materially alter the microbial community inferred from the same underlying sequencing data, yet head-to-head comparisons are often confounded by mismatched reference databases or preprocessing. We benchmarked three widely used 16S rRNA pipelines—MetagenApp, mothur, and QIIME 2—under identical reference (SILVA 138.2 NR99), primer-trimming, and filtering conditions, using 36 paired-end MiSeq V3–V4 samples (pharyngeal and nasal, 1,519,852 raw reads). MetagenApp and mothur inferred features via 97%-similarity clustering (VSEARCH), while QIIME 2 used DADA2 ASV inference. MetagenApp produced 12,599 features (69.9% reads retained) versus 7,552 OTUs for mothur (68.2%) and 1,332 ASVs for QIIME 2 (66.5%). MetagenApp was substantially faster and more memory-efficient (14 min 55 s, 6.8 GB RAM) than QIIME 2 (20 min, 14.1 GB) and especially mothur (2 h 24 min, 33.8 GB). Despite differences in feature counts and alpha diversity—clustering-based pipelines recovered higher raw richness and Chao1—community structure was highly concordant across pipelines (Mantel r = 0.94; PERMANOVA: sample identity R² = 88.9% vs. pipeline choice R² = 1.90%), and phylum-level taxonomic profiles remained broadly consistent, with unclassified reads below 0.5% for all three pipelines. Our results show that biological signal dominates over pipeline-specific effects, while highlighting real trade-offs in computational cost and feature resolution.
Co-authors: None
Contact Attendee
Warning: Attempt to read property "user_email" on string in /home/1276969.cloudwaysapps.com/ydbgzhdjeq/public_html/wp-content/plugins/my-conference-now/functions.php on line 2826