dbcan-nf: a Nextflow pipeline for accurate CAZyme annotation in metagenomes

Motivation CAZymes (carbohydrate-active enzymes) are crucial for carbohydrate degradation in any ecological environments. However, CAZyme annotation is not easy in whole genome shotgun (WGS) metagenomes, due to the lack of accurate, scalable, robust, and reproducible workflows. We previously published dbCAN as a web server and associated standalone tool run_dbcan that performs CAZyme annotation with assembled genomes as the input but not with raw WGS reads. Results Here, we present dbcan-nf, an accurate and scalable Nextflow pipeline integrating run_dbcan as a component step, for automated CAZyme annotation in metagenomes. Dbcan-nf supports multiple analytical routes including individual sample assembly (IA), multi-sample co-assembly (CA), assembly-free (AF), and subsampling (SS), and accepts both short and long reads. Using five benchmark datasets, we showed that dbcan-nf (IA and AF routes) had a much higher accuracy (recall and F1 score) and detected more CAZyme families than the competing tool Cayman. Among the different routes, dbcan-nf-IA produced the most accurate CAZyme annotation, especially when the sequencing depth is sufficient (e.g., >1.0 Gb). However, as a read assembly-based method, dbcan-nf-IA had a higher computational cost than Cayman. Lastly, dbcan-nf leverages Nextflow-supported High-Performance Computing (HPC) execution for workflow-level scalability through concurrent processing of multiple samples. Availability and implementation https://github.com/bcb-unl/dbcan-nf .