Dropping Gene Models

[alyssa....@uconn.edu]

Hi, 

I was wondering if evidence modeler will drop gene models if they are only provided in one type of evidence. I also wanted to confirm that evidence modeler sometimes can split gene models and combine gene models from different types of evidence. 

I'm trying to resolve gene fragmentation (from braker) using protein and transcript evidence (from other sources), and have been facing difficulty.

Thanks!

[Brian Haas]

Hi Alyssa,

responses below

On Thu, Jul 12, 2018 at 2:36 PM <alyssa....@uconn.edu> wrote:

Hi, 

I was wondering if evidence modeler will drop gene models if they are only provided in one type of evidence.

Yes.  In the case there are multiple ab initios, any place where each ab initio does not predict a gene, it's contributing a 'not a gene' type of score, and this (intergenic) score could end up out-weighing the alternative (genic) score.  

 

I also wanted to confirm that evidence modeler sometimes can split gene models and combine gene models from different types of evidence. 

Yes.  Each prediction is broken down into its component parts (exons, introns, inter-genic) and scored according to the weight assigned.  The predictions EVM generates involves the combination of these parts that gives the highest cumulative score and are properly compatible with each other (ie. exons link together that are in-phase and connected by an intron, etc).

 

I'm trying to resolve gene fragmentation (from braker) using protein and transcript evidence (from other sources), and have been facing difficulty.

If you use protein and transcript alignment evidence with EVM, just be sure that they're proper spliced alignments.  Be sure to look at the formatting of the example data provided.

best,

~brian

 

Thanks!

--
You received this message because you are subscribed to the Google Groups "EVidenceModeler-users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to evidencemodeler-...@googlegroups.com.
To post to this group, send email to evidencemo...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/evidencemodeler-users/0378ce5c-9417-4734-a15d-1b8e69e97aac%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

--

--
Brian J. Haas
The Broad Institute
http://broadinstitute.org/~bhaas

 

[alyssa....@uconn.edu]

Thank you!!

[Brian Haas]

(I forgot to CC the group)

Will evidence modeler drop gene models specifically if they are only in an abinitio prediction and not in protein or transcript evidence?

No, in this case, it depends on the balance among the ab initios that predict in that region vs. those that do not, and their relative weights.

 

[alyssa....@uconn.edu]

I am trying to use EVM to resolve fragmentation in the annotation of a conifer species (douglas fir). I am inputting a Braker gff with 45334 genes, a protein alignment gff with 33884 genes, and transcripts from Maker (13794 genes) and Illumina (23410 genes) which include partial genes. EVidence Modeler is only returning 25902 genes and has shortened many introns and genes. I have attached an excel file with comparisons of statistics about genes from both files, but I am wondering why so many of my genes are being removed (20357), without overlapping any of the new EVidence Modeller genes. I am also wondering why it seems many of my genes are being shortened. I am currently weighting my genes 4 and proteins and transcripts 6. Is it necessary to use PASA to get EVM to work for my data?

Thank you for any help,

-Alyssa