fastnb: A high-performance C++ Naive Bayes classifier for SILVA-based 16S rRNA taxonomy

We present fastnb , an open-source C + + 17 reimplementation of the scikit-learn Multinomial Naive Bayes classifier used in the QIIME 2 feature-classifier plugin for 16S rRNA amplicon taxonomy assignment. By replacing Python-level inference with a native C + + engine employing batch matrix multiplication, transposed memory layout for SIMD auto-vectorization, and single-precision arithmetic, fastnb achieves a 13–21 × speedup over scikit-learn on eight threads while maintaining 99.85% genus-level classification agreement across six clinical patient samples (1993 ASVs total). The software is provided as a pip-installable Python package with pybind11 bindings and requires no changes to existing QIIME 2 workflows.