The Naïve Bayes classifier++ for metagenomic taxonomic classification-query evaluation

Area of Science:

• Bioinformatics
• Computational Biology
• Metagenomics

Background:

• Metagenomic analysis is crucial for understanding microbial communities.
• Existing tools like Kraken2 face challenges with memory and database size.
• The Incremental Naive Bayes Classifier (NBC++) was developed to address these limitations.

Purpose of the Study:

• To evaluate the query performance of NBC++ across various parameters.
• To compare NBC++ with Kraken2 in terms of efficiency and accuracy.
• To identify factors influencing NBC++ performance, including database depth and k-mer size.

Main Methods:

• Testing NBC++ with different canonicality settings, k-mer sizes, and database depths.
• Analyzing the impact of input sample data size on query performance.
• Benchmarking NBC++ against Kraken2 for training time, memory usage, and query speed.

Main Results:

• NBC++ demonstrates competitive superkingdom profiling using smaller databases.
• NBC++ requires less training time and memory compared to Kraken2, but has longer query times.
• Both NBC++ and Kraken2 performance improve with increased database depth, though capturing full biodiversity, especially viruses, remains difficult.
• Enhancements in NBC++ include canonical k-mer storage for reduced memory footprint and optimized memory allocation for faster analysis.

Conclusions:

• NBC++ provides an efficient alternative for metagenomic profiling, particularly in resource-constrained environments.
• Database depth significantly impacts the performance of both NBC++ and Kraken2.
• Further development is needed to fully capture microbial diversity in metagenomic datasets.