-
Notifications
You must be signed in to change notification settings - Fork 88
Description
Dear funannotate team
Firstly thanks for the fantastic work on this pipeline. I have been using funannotate as part of my genome annotation work in conjunction with other tools to obtain different forms of evidence (e.g. TAMA pipeline using isoseq data, GeMoMa based on homology, and other abinitio predictors).
Funannotate predict has the option 'other_gff' to provide additional annotations which are then assigned the 'OTHER_PREDICTION' label in the weights file that the funannotate script generates and provides to EVM. From what I understand though, the annotations labeled 'OTHER_PREDICTION' are used differently by EVM to those labeled 'ABINITIO_PREDICTION', in that the former do not contribute to the identification of intergenic regions.
Therefore, if some of my 'other' gffs come from another abinitio pipeline / tool such as BRAKER or HELIXER, it would be good to have the option to classify them as such when running funannotate predict so that EVM can treat them as intended. This could be set along with the weights, e.g.:
--other_gff helixer.gff:3:abinitio tama.gff:10:other
I have found that running EVM separately and labelling additional abinitio predictions as ABINITIO_PREDICTION in the weights file achieves higher BUSCO completeness than when running funannotate predict with the same annotations specified as --other_gff. At the moment therefore I am running EVM separately in my annotation work, though funannotate predict is still serving as a very convenient wrapper for all the prior steps (exonerate, augustus, genemark, and prepping inputs for EVM). It would be great to have the additional flexibility with funannotate predict as suggested above, so those of us wanting to use additional abibitio predictions don't need to run EVM separately to obtain the best results.
Cheers
James