nextgenusfs
commented
4 years ago I would do the following:
-
Softmask repeats in genome assembly;
funannotate maskuses RepeatModeler/RepeatMasker, although you can use any other method as well. doing this step first allows you to work with the same genome file throughout the pipeline. -
Use RNA-seq data for genome guided transcriptome assembly and training of ab initio gene predictors.
funannotate trainwill run your steps 1-3 automatically, i.e. assuming you have PE RNA seq from 3 conditions you would run:funannotate train -i softmasked.genome.fa -o annotate \ --left rna1_R1.fq.gz rna2_R1.fq.gz rna3_R1.fq.gz \ ---right rna1_R2.fq.gz rna2_R2.fq.gz rna3_R2.fq.gz \ --stranded RF --species "Diatom species" --cpus XX --max_intronlen 10000This command will run Trimmomatic on your RNA-seq reads, combined RNA-seq data, run in-silico normalization, align normalized reads to genome, assembly into transcripts using Trinity, align transcripts to genome using PASA, choose best gene model for each alignment using Transdecoder.
-
The "beauty" of running the defined method in
funannotate trainis that all of the data will be automagically used byfunannotate predict. So you can then run the following to continue training gene predictors and predict best consensus gene models.funannotate predict -i softmasked.genome.fa -o annotate --busco_db protists \ --organism other -- max_intronlen 10000 --species "Diatom species" --cpus XXSome users have provided feedback that
CodingQuarrydoesn't work well on non-fungal genomes, I haven't disabled it yet for non-fungi genomes, however, if you see thatCodingQuarryis predicting 2X the number of genes as the other gene predictors, you can turn it off by adding--weights codingquarry:0to thefunannotate predictcommand. -
Since you have RNA-seq data, you can add alternative transcript and add UTRs using
funannotate update, again pipeline will re-use existing data.funannotate update -i annotate --cpus XX -
Finally you can add functional annotation to your annotated genome:
funannotate annotate -i annotate --cpus XXAt the end of each script, there will be a "suggested command" of what you might want to run next. Good luck! -Jon
jolespin
I'm trying to do following to assemble a diatom:
(1) Kneaddata/Trimmomatic for my transcript reads; (2) Mapping the reads to my unmasked genomes; (3) Using RNA-Spades for the transcript assembly using the reads from [2]; (4)
funannotate trainfor each of my libraries ( I have a lot); (5) Use the following to predict genes usingfunannotate predict: (5a) All of my training files from [4] ( I believe these will be gff3 files fromfunannotate train)--pasa_gff; (5b) trusted proteins from the nearest organism--protein_evidence; (5c) A merged map file for all of my transcriptomesMy questions:
(1) Should the above protocol work? (2) Can I use multuple --rna_bam? (3) Does my --rna_bam need to be aligned to the repeat masked assembly? (4) Does my
funannotate trainstep need to be on the repeat masked assembly?