Query Regtarding Funannotate/MySQL

[ufaroooq]

Hi,

I am posting it here to just get your openion before diving into the process.

Previously i had 10 samples which i needed to annotate, so i used Funannotate v1.8.17 the conventional way by installing it via conda and also installed all required lincenced programs. and it worked. took too long time on some steps specifically related to PASA.

Now i have to annotate around 115 genomes too. all are different. and if complete pipeline takes 4-days then overall it will take way too much time/days. Even using HPC where i can book upto 10 nodes, it will still take alot of time.

Things i am thinking of doing.

• During funannotate train raw reads nortmalization taked around 4 hrs for me, but i checked those files are generated without any kind of incorporation of genome-fasta, so my guess is that i can re-use those files. (funannotate does accept pre-normalized files too) Question here: Will this be a good thing to do or it can introduce any biseness?
• secondly, the most time taking step is PASA as it uses SQL-lite which is single-threaded. Funannotate documentation does suggest a way to use MySQL so PASA can work on multiple threads, but mySQL becomes problomatic on HPC due to security.

For that i was thinking, to create my own Docker/singularity container, where i install complete funannotate and set it up to use MySQL instead of SQL-lite. I am not sure if the pre-build container from OG-developer is based on MySQL.

I just need your thoughts, Will this be a good approach ?, Is it possible to do. As my guess is that i have to install all tools and dependencies manually without conda.
Or if there is already a container set up with MySQL available ?

YOur input/thoughts will be helpfull.

Regards

[hyphaltip]

yes I often normalize once and then have those files symlinked into all the other - eg when I annotate hundreds of strains of same species I sometimes train once, then re-use the trimmed and normalized files throughout by making a symlink to the master one. I don't remmeber which project has this right now but you can do that.

yes setting up a mysql local that can just run during the train step if you want, you can re-use it for the update step if you save the db in common place though sometimes i just have this run on scratch partition and throw it away after train step.

In this example I setup a singularity mysql run before I start the train step, so it runs on same node as the training job, if you want to have multiple runs (eg lots of genomes being annotated) then you be creating a separate server for each run, or you can pre-launch a server and as long as there aren't too many running at same time it can handle it.

https://github.com/stajichlab/PopGenomics_Coccidioides_2025/blob/main/Assembly/pipeline/02-annotate/02_RNAseq_train.sh