Can OMArk classifications be manually adjusted when the default settings fail?

[HanShiling]

Hello,

Thank you for developing such a useful and user-friendly tool. In most cases, OMArk is very sensitive and has helped me identify and remove contaminant sequences from proteomes.
However, I have noticed that for a few eukaryotic species, OMArk does not seem to classify them correctly.

I would like to ask: when the default OMArk parameters cannot correctly assign the sampled and closest lineages, is there a way to manually adjust or specify them?

Below are several examples I encountered. In Example 1, the proteome I analyzed, SRR8610319, corresponds to the species Nannengaella melleum, which belongs to Amoebozoa/Protozoa. However, OMArk classified it as Fusarium, which belongs to Fungi.
Could you please advise what would be the recommended way to correct or refine the classification in such cases?

Example 1: Nannengaella melleum (SRR8610319, Physaridae, Physarida, Myxogastrea, Amoebozoa, Protozoa)
cat SRR8610319_cleaned_strict/SRR8610319_cleaned_strict.tax

Sampled
Fusarium
Closest
Gibberella moniliformis (strain M3125 / FGSC 7600)

Example 2: Phialemonium thermophilum (GCA_042366435.1, Cephalothecaceae, Cephalothecales, Sordariomycetes, Ascomycota, Fungi)
cat GCA_042366435.1_cleaned_strict/GCA_042366435.1_cleaned_strict.tax

Sampled
Sordariales
Closest
Sordariales

Example 3: Sakaguchia melibiophila (NMDC60198298, Sakaguchiaceae, Sakaguchiales, Cystobasidiomycetes, Basidiomycota, Fungi)
cat NMDC60198298_cleaned_strict/NMDC60198298_cleaned_strict.tax

Sampled
Pucciniomycotina
Closest
Pucciniomycotina

Example 4: Cystobasidiopsis lactophila (GCA_001599975.1, Chionosphaeraceae, Agaricostilbales, Agaricostilbomycetes, Basidiomycota, Fungi)
cat GCA_001599975.1_cleaned_strict/GCA_001599975.1_cleaned_strict.tax

Sampled
Pucciniomycotina
Closest
Pucciniomycotina

Although this issue affected fewer than 5% of the approximately 1,000 genomes I analyzed, I would still appreciate any suggestions you may have on how to address it.

If it would be helpful, I would be happy to share the proteome files I used, together with the corresponding OMArk output files, by email.

Thank you very much for your time and help!

Best,

Shiling

[YanNevers]

Dear @HanShiling ,

Thank you for your kind words!
You can directly submit the taxonomic identifier of your species using the -t option from OMArk. As in:
omark -f example_data/UP000005640_9606.omamer -t 9606 -d LUCA.h5 -o example_data/omark_output

It will then not use this placement for the following analysis rather than the automatic placement.

Regarding the example you provided, Example 1 is the most concerning to me, because it appears like it does not successfully place the genes from the expected species as belonging to Amoebozoa. I would be interested to see the OMArk output files to better understand this.

As for the other examples, it seems they are mainly due to a lack of sampling in the OMArk database.
To my understanding, Example 2 is OMArk placing this species as a sister group from the actual taxon, because representative from this taxon are not known to it .
Example 3 and 4 place in a correct taxon, although not that specific, likely because more precise clades are not present in OMArks database.

Although this should get better as the OMA database increase is size, lacking specificity is unavoidable. Using the -t option will make sure the taxon OMArk chose does include the species of interest.

I hope it answers your questions.

Best wishes,
Yannis

[HanShiling]

Dear @YanNevers

Thank you for your reply. I reran Example 1 using the parameters you recommended. The specific command I used is as follows:

omark -f SRR8610319_longest_cleaned_strict.omamer -t 289959 -d LUCA.h5 -o SRR8610319_longest_cleaned_strict/ -e ~/.local/share/ete/taxa.sqlite

The tax results are as follows:

cat SRR8610319_longest_cleaned_strict.tax

Sampled
Evosea
Closest
Evosea

The summary results are as follows:

cat SRR8610319_longest_cleaned_strict.sum
#The selected clade was Evosea
#Number of conserved HOGs is: 946
#Results on conserved HOGs is:
#S:Single:S, D:Duplicated[U:Unexpected,E:Expected],M:Missing
S:378,D:510[U:489,E:21],M:58
S:39.96%,D:53.91%[U:51.69%,E:2.22%],M:6.13%
#On the whole proteome, there are 73340 proteins
#Of which:
#A:Consistent (taxonomically)[P:Partial hits,F:Fragmented], I: Inconsistent (taxonomically)[P:Partial hits,F:Fragmented], C: Likely Contamination[P:Partial hits,F:Fragmented], U: Unknown
A:17512[P:4216,F:4709],I:9883[P:3688,F:2977],C:12099[P:4037,F:4855],U:33846
A:23.88%[P:5.75%,F:6.42%],I:13.48%[P:5.03%,F:4.06%],C:16.50%[P:5.50%,F:6.62%],U:46.15%
#From HOG placement, the detected species are:
#Clade NCBI taxid Number of associated proteins Percentage of proteome's total
Acanthamoeba castellanii str. Neff 1257118 5704 7.78%
#Potential contaminants:
Gibberella moniliformis (strain M3125 / FGSC 7600) 334819 12227 16.67%
Metakinetoplastina 2704647 924 1.26%

I was wondering whether the difficulty in classifying this sample might be due to the presence of contaminating sequences, or perhaps to insufficient proteome quality.

Best wishes,

Shiling