Scalable Functional Classification Pipelines for Phytozome Plant Genes: InterProScan and PMN implemented in Nextflow

Phytozome, the Joint Genome Institute’s portal for Comparative Plant Genomics, hosts ~60 assembled and annotated plant genomes, drawn from a broad range of Viridiplantae species with relevance to biofuels, climate change adaptation, and plant evolution. Approximately two-thirds of these genomes were assembled and/or annotated at the JGI, with the remainder acquired from external sources. All gene sets are processed through the Phytozome functional classification pipeline, which includes PFAM/KOG/PANTHER/KEGG and related components.

Previous attempts to use InterProScan5 and the PMN pathway assignment workflow for functional classification of gene data sets released through Phytozome have had limited success. Few groups outside of EBI (InterPro’s sponsoring institution) have been able to run InterProScan5 at scale on typical bioinformatics computing clusters implementing SGE-like schedulers. Similarly, the PMN family of assignment modules (E2P2, Pathologic, SAVI) requires significant integration and parallelization work to handle updates to ~60 genomes in an automated workflow. Therefore many groups continue running older versions of InterProScan and restrict their pathway classifications to the genomes available directly from PlantCyc.

We report here on the use of the Nextflow system to implement InterProScan5 and PMN workflows for scalability and portability. We now have an implementation of InterProScan5 that runs on both single workstations and SGE-based compute clusters, and an integrated pipeline for pathway assignment, both of which handle job parallelization and dependencies in a lightweight framework. We report on the structure and performance of these pipelines, their availability, and plans for Nextflow implementations of additional analysis pipelines in Phytozome.