KOBAS evaluation

[mpoelchau]

It is worth exploring whether there are better alternatives to insect KEGG pathway annotation than KOBAS since when we developed the pipeline ~5 years ago.

General considerations:

• have other databases/entities started annotating Refseq proteins with KEGG pathways? (e.g. KEGG itself; uniprot)
  • eggnog is often wrong wrt finding orthologs
• what software programs exist that we could use, instead? Create list with pros and cons.
• should we explore using flybase pathway annotations? (are they free; machine readable?)? Wormbase? Alliance?

Software considerations:

• Is the software maintained?
• Is it already used/adopted by the bioinformatics community?
• Is the pathway annotation transferred directly from the entity that was curated to the new protein, or are there intermediate steps?

Other considerations:

• InterProScan is not adding Reactome data anymore via AgBase; script to add it is broken. (How does Interproscan actually assign the Reactome data to proteins?)

[amcooksey]

Interproscan script fixed. Now adding Reactome annotations again.

[amcooksey]

Flybase pathways:
downloaded this file:
pathway_group_data_fb_2024_04.tsv

There are 17 parent pathways.
There are 73 unique entries in column 2 (I think those are the roles in the pathways).
There are 681 annotated genes

They don't cover many genes. Pathways seem pretty basic. I don't think there is anything here we can't get somewhere else.

We could probably anntoate to them in the same way we do GOanna/Kofamscan--BLAST/HMM match the genes to drosophila homologs and pull those annotations across.

[amcooksey]

Things I found to do KEGG annotations:
Kofamscan/ClusterProfiler combo:

• kofamscan is the standalone version of KofamKoala from KEGG. Pulls KO number for each FASTA protein.
• clusterprofiler is a Bioconductor tool with a KEGG mapper option. This takes the KO number and pulls the corresponding pathway annotations.

KEGG API:
Can pull a variety of tables that get the annotations for any species in KEGG. For other species we would still need to find homologs in KEGG to pull annotations from.