PRANK aligner problems - Githubissues

davidemms / OrthoFinder

Phylogenetic orthology inference for comparative genomics

https://davidemms.github.io/

GNU General Public License v3.0

716 stars 189 forks source link

PRANK aligner problems #601

Open fka21 opened 3 years ago

fka21 commented 3 years ago

Hello David,

Thank you for creating and continuously developing this great tool! I wanted a run with PRANK aligner with the run fails even though the test alignment runs successfully. Here is the log of it: prank.txt Also the config file: config.txt

Also I have am curious about the species tree interference. I have read a few github issues regarding it, but could not find an answer to my question. To my experience when providing a guide species tree without branch lengths OrthoFinder returns a species tree without branch lengths also. Is this normal? I thought OrthoFinder uses the user supplied species tree as guide and it will estimate a new tree based on the guide tree and the data provided. Thank you very much in advance for the help!

Best, Ferenc

davidemms commented 3 years ago

Hi Ferenc

From the look of the prank output if doesn't write the alignment to the name specified by "-o=OUTPUT" but instead to 'OUTPUT.best.fas' so try adding this to the config entry:

"ouput_filename": "OUTPUT.best.fas"

However, you may still have issues with prank. When I tried it on the Example Dataset a number of alignments failed. Prank seemed to run fine if run on those files individually, so I wonder if it writes to some temporary files that get mixed up between different instances of prank?

Good question re the species tree. No, OthoFinder only requires the topology of the species tree. It only needs the tree for a couple of uses:

identifying the outgroup set of genes in each gene tree in order to root the tree
identifying all the clades in the species tree so it can return a set of HOGs at that clade

Neither of these require branch lengths.

All the best David

fka21 commented 3 years ago

Thank you for the reply, in that case I might fall back to some of the default aligners.

Regarding the species tree do you know why it might happen? I would need a species tree with branch lengths for downstream analysis. If I provide a guide tree it returns a correct topology of the species tree, but with branch lengths of 1. If I let OrthoFinder estimate the tree on it's own it returns a tree with bad topology. What do you think I could do to overcome this?

davidemms commented 3 years ago

For the species tree, have you provided enough data? How many species do you have and what is the average number of sequences per species? You could using something like iqtree tree to infer branch lengths for a pre-defined topology:

OrthoFinder to infer it using the "-M msa" option, this will give you a concatenated MSA in "MultipleSequenceAlignments/SpeciesTreeAlignment.fa"
Run IQTREE with the MSA and your tree topology using "-te" option to fix the tree.

Best wishes David

fka21 commented 3 years ago

I am using 43 species with proteomes sizes from ~13.000 to ~60.000 peptides. Most of the proteomes come from ensembl with some coming from de novo transcriptome assemblies (redundancy of these has been shrunken using evigene pipeline).

Unfortunately OrthoFinder did not provide MSA used for species tree estimation when a guide tree. I could find one in the run without guide tree though. I could try to use this one with iqtree. Thank you for the tip!