Intronarrator is a collection of Python and shell scripts that predicts introns using RNA-seq data, then uses the program AUGUSTUS to do gene prediction, using its "intronless" model.
The motivation for this is that AUGUSTUS's intron model leads to poor prediction of extremely short (mostly 15 bp) introns of heterotrichous ciliates, despite having an intron length model designed to accommodate extreme variation in intron lengths observed across eukaryotes. After trying a number of options to alter this model, we were unable to accurately predict introns - leading, more often than not, to predicted introns being incorrect ("faketrons"). RNA-seq data, on the other hand, may be very extensive -- providing deep coverage of most genes. Thus, it is possible to predict the introns directly from RNA-seq data, and leave the rest of the gene prediction to AUGUSTUS.
The approach is illustrated below (semi-transparent green indicates
where a gene to be predicted is located):
The following programs need to be installed: AUGUSTUS, Infernal, tRNAscan-SE version 2.0.
Code can be obtained from github:
git clone https://github.com/Swart-lab/Intronarrator/
An Intronarrator evironment with all the necessary depencies can be installed from the included env.yml with conda:
cd Intronarrator conda env create -f env.yml conda activate Intronarrator
intronarrator.sh needs to be copied to the working directory. After this it needs to be edited and properly setup before running, including setting the paths to AUGUSTUS configuration files. You should also decide on the number of parallel processes to run.
Helper scripts for producing stranded BAM files, and generation of "hints" for AUGUSTUS from the BAM files are included in the directory "helper_scripts". These should be copied and modified according to need before use.
Reads with multiple introns currently are ignored - these still need to be properly accommodated. Therefore be warned: this software is currently inappropriate for organisms with high densities of introns and frequent alternative splicing (e.g. most metazoans)!