search

Gaius-Augustus / Tiberius

deep learning gene finder
MIT License
33 stars 3 forks source link

[Question] Can this be used for metagenome-assembled genomes targeting protists and fungi? #5

Open jolespin opened 1 month ago

jolespin commented 1 month ago

I’m looking for gene prediction alternatives to MetaEuk that are less resource intensive for my VEBA software suite (https://github.com/jolespin/veba). With a cursory glance at the repo, it looks like Tiberius is tested on higher eukaryotes but how is the usage (and representation in training data) on diverse protists (eg algae, amoeba, ciliates, etc) and fungi? Also, in metagenomics we would also have fragmented genomes and I’m not sure if this would affect performance. I would not know the lineage of the genomes a priori since they are environmentally acquired.

From the other issues it looks like a conda version will become available soon that I can test out.

KatharinaHoff commented 1 month ago

I have, in fact, trained Tiberius on diatoms as an example for exactly this purpose, but the model is not published, yet. I would not recommend to use the mammalian model on protists, etc.

I am currently working under pressure of several deadlines, but I will be happy to discuss with you at some point in October or November. Send me an e-mail? katharina.hoff@uni-greifswald.de

jolespin commented 1 month ago

Yea that sounds great. I'll try to ping you in November after the e deadlines. Ideally I would like to use this for general microeukaryotic gene modeling but more specifically right now I need it for marine microeukaryotes (many diatoms, algae, and fungi).