Skip to content

Latest commit

 

History

History
119 lines (83 loc) · 4.04 KB

pipelines.md

File metadata and controls

119 lines (83 loc) · 4.04 KB
Raw

Annotation Pipelines

The default pipeline is suitable for most purposes. However, dammit has several alternative workflows that either reduce the number of databases and tools run (the quick pipeline) or annotate with larger databases, such as UniRef90 ( full pipeline).

Note: To use these pipelines, first run dammit run databases to make sure the relevant databases are installed. Then you can proceed with dammit run annotate

Default

By default, dammit runs the following:

  • default:
    • buscoquality assessment
    • transdecoder ORF prediction
    • shmlast to any user databases
    • hmmscan to Pfam-A
    • cmscan to Rfam
    • LAST mapping to OrthoDB and Swiss-Prot

The databases used for this pipeline require approximately ~18GB of storage space, plus a few hundred MB per busco database. We recommend running this pipeline with at least 16GB of RAM.

code to run this pipeline:

dammit run databases --install
dammit run annotate

If specifying a custom location for your databases, add --databases-dir /path/to/dbs

Alternative annotation pipelines:

quick pipeline

  • quick (--pipeline quick):
    • buscoquality assessment
    • transdecoder ORF prediction
    • shmlast to any user databases

The quick pipeline can be used for running a minimal annotation run: BUSCO quality assessment, ORF prediction with transdecoder, and shmlast to map to any user databases. While this pipeline may require less database space, we still recommend running with 16G of RAM, especially if mapping to a user-provided protein database.

code to run this pipeline:

dammit run databases --install --pipeline quick
dammit run annotate --pipeline quick

If specifying a custom location for your databases, add --databases-dir /path/to/dbs

full pipeline

warning: time and resource intensive!

The full pipeline starts from the default pipeline and adds a mapping database, UniRef90.

UniRef90 is a set of UniProt sequences clustered by >=90% sequence identity. UniRef allows a searching to a larger set of sequence records while hiding redundant sequences. See the UniRef documentation for more.

  • full (--pipeline full):
    • busco quality assessment
    • transdecoder ORF prediction
    • shmlast to any user databases
    • hmmscan to Pfam-A
    • cmscan to Rfam
    • LAST mapping to OrthoDB, Swiss-Prot, and UniRef90

As of fall 2020, the UniRef90 fasta is 26G (gzipped).

code to run this pipeline:

dammit run databases --install --pipeline full
dammit run annotate --pipeline full

If specifying a custom location for your databases, add --databases-dir /path/to/dbs

nr pipeline

warning: REALLY time and resource intensive!

nr is a very large database consisting of both non-curated and curated database entries. While the name stands for "non-redundant", this databse is no longer non-redundant. Given the time and memory requirments, nr is only a good choice for species and/or sequences you're unable to confidently annotate via other databases.

  • nr (--pipeline nr):
    • busco quality assessment
    • transdecoder ORF prediction
    • shmlast to any user databases
    • hmmscan to Pfam-A
    • cmscan to Rfam
    • LAST mapping to OrthoDB, Swiss-Prot, and nr

As of fall 2020, the nr fasta is 81G (gzipped).

code to run this pipeline:

dammit run databases --install --pipeline nr
dammit run annotate --pipeline nr

If specifying a custom location for your databases, add --databases-dir /path/to/dbs

Note: Since all of these pipelines use a core set of tools, and since dammit uses snakemake to keep track of the files that have been run, dammit will not rerun the core tools if decide to alter the pipeline you're running. So for example, you could start by running a quick run, and later run default if desired. In that case, dammit would run only the new annotation steps, and reintegrate the relevant outputs into new dammit gff3 and annotated fasta files.