swarbred
commented
3 years ago Hi @francicco
Can you be clear by what you mean by "nested" spliced models entirely within introns will not generally be removed by the default configuration, if you are referring to models with partial exon overlaps then there are a few place in the config that can be adjusted
see
[pick.clustering]
# Parameters related to the clustering of transcripts into loci.
# - cds_only: boolean, it specifies whether to cluster transcripts only according to their CDS (if present).
# - min_cds_overlap: minimal CDS overlap for the second clustering.
# - min_cdna_overlap: minimal cDNA overlap for the second clustering.
# - flank: maximum distance for transcripts to be clustered within the same superlocus.
# - remove_overlapping_fragments: boolean, it specifies whether to remove putative fragments.
# - purge: boolean, it specifies whether to remove transcripts which fail the minimum requirements check - or whether to ignore those requirements altogether.
# - simple_overlap_for_monoexonic: boolean. If set to true (default), then any overlap mean inclusion in a locus for or against a monoexonic transcript. If set to false, normal controls for the percent
# - max_distance_for_fragments: maximum distance from a valid locus for another to be considered a fragment.
cds_only = false
min_cds_overlap = 0.2
min_cdna_overlap = 0.2
purge = true
flank = 200
simple_overlap_for_monoexonic = trueIf the "missing" genes are mono exonic switching simple_overlap_for_monoexonic = false will allow some mono exonic genes to be retained with a partial overlap.
You can also adjust min_cds_overlap = 0.2, min_cdna_overlap = 0.2 to allow a greater overlap (for both mono and multi exonic genes)
Changing cds_only = false to cds_only = true will define clusters based on cds structure overlap i.e. to belong to the same gene multi exonic transcripts needs to share a coding splice site. That will mean that genes which share only splice junctions in UTRs will be output as separate genes.
There may be other reasons why the transcripts you regard as missing are removed i.e. they may not meet the requirements in the scoring file (see scoring file requirements:) or could be even removed during the prepare stage.
If your nested models share splice sites with another selected model then these may be excluded due to the as_requirements (see the scoring file).
If these are mono exonic models then they may be being filtered by the not_fragmentary expression (scoring file) you can either adjust this expression or change how fragments are handled in the pick.fragments section of the configuration
[pick.fragments]
# Parameters related to the handling of fragments.
# - remove: boolean. Whether to remove fragments or leave them, properly tagged.
# - max_distance: maximum distance of a putative fragment from a valid gene.
# - valid_class_codes: which class codes will be considered as fragments. Default: (p, P, x, X, i, m, _). Choices: '_' plus any class code with category 'Intronic', 'Fragment', or 'Overlap'.
remove = true
max_distance = 2000
valid_class_codes = ["p", "P", "x", "X", "i", "m", "_", "e", "o"]The class codes https://mikado.readthedocs.io/en/latest/Usage/Compare.html#class-codes define what can be regarded as a fragment removing "i" would exclude intronic mono exonic transcripts from this definition.
If you can show a browser screen shot to show the kinds of models that are being removed that might help me to advise. But if you are losing mono-exonic loci then removing "i" from the fragments class codes and switching simple_overlap_for_monoexonic = false may achieve what you want.
I'm also going to give a shameless plug for the workshop we are organising for April next year that may by of interest https://www.earlham.ac.uk/genome-annotation-workshop-2021
Best David
francicco
lucventurini
Hi,
I'm trying to find a way to tell mikado not to throw away nested genes. Looking at the mikado tutorial, I still haven't find anything. which are the parameters I should change or introduce?
Thanks a lot F