From defaults to databases: parameter and database choice dramatically impact the performance of metagenomic...

Area of Science:

• Microbiology
• Bioinformatics
• Computational Biology

Background:

• Taxonomic classification of microbiome reads is crucial for analyzing microbial communities.
• Kraken2 (k-mer-based) and MetaPhlAn 3 (marker gene-based) are leading tools for this task.
• Discrepancies exist between these tools in classifying human-associated and environmental metagenomes.

Purpose of the Study:

• To compare the performance of Kraken2 and MetaPhlAn 3 for metagenomic taxonomic classification.
• To evaluate the impact of tool-parameter-database choices on classification accuracy.
• To determine the optimal tool for different research scenarios and resource constraints.

Main Methods:

• Utilized simulated and mock metagenomic samples to assess classification accuracy.
• Compared Kraken2 and MetaPhlAn 3 using various datasets (human-associated, environmental).
• Examined the influence of different tool parameters and databases on taxonomic assignments.

Main Results:

• Significant discrepancies were observed in read classification proportions and identified species between Kraken2 and MetaPhlAn 3.
• Kraken2 demonstrated superior precision, recall, and F1 scores, with diversity measures closer to known compositions.
• MetaPhlAn 3 required fewer computational resources, making it more accessible for some researchers.

Conclusions:

• There is no universal 'best' tool; the optimal choice depends on the research question and available resources.
• Kraken2 offers higher performance but may be computationally prohibitive.
• Careful consideration of tool, parameters, and database is essential for accurate metagenomic analysis.