Finding Orthologs in Unannotated Genomes

Question

[triplee0305]

Hi all,

I try to identify the ortholog of a particular gene across genomes of many species. However, most of the genomes I am working with are unannotated.

What tools or workflows would you recommend for finding orthologs in unannotated genomes?

From my research, tools like OrthoMCL seem to require annotated genes or, at the very least, an identification of coding regions. Given that I don't have this information, I am unsure how to proceed.

I was considering using tools like MiniMap2 to align the gene directly to the genomes, as it supports long-read alignments. Would this approach work? I assume this approach would tell me whether this gene is present or not in the genome?

Could anyone provide guidance or suggest alternative tools/methods that might work in this scenario?

Thank you!

Answer 1

[shelkmike]

You can take the protein of your gene and align it to the genomes by Miniprot. It is a splicing-aware protein-to-genome aligner.

[triplee0305]

Thank you for the suggestion!

I have tried it on more than 300 genomes now. Only several genomes got positive result. It seems hard to judge whether is false negative (because I use the default settings for all species), or it is really missing from a particular species.

Answer 2

[Philipp Bayer]

I think what you want is PaGeSearch: http://genome.cshlp.org/lookup/pmidlookup?view=long&pmid=38858086

https://github.com/Sohyoung/PaGeSearch

It aligns candidate genes with an unannotated genome via MMSeqs2/exonerate, then uses AUGUSTUS and the candidate genes for de novo prediction in the hit region, then filters the output using a pretrained artificial network.

PaGeSearch can find 67 genes in the O. sativa’s genome [...] in 10 min on 4 CPU.

[triplee0305]

This is indeed very cool! Unfortunately, I am working on insects, which is not yet included in their taxonomic classes. But thank you for this information!