Unexpected Low % Associated Query Proteins within the Main Species/Clade

[glmaala]

I'm using the software to assess protein sets of multiple molluscan species.

Correct me if I'm wrong, but I've inferred from the values obtained that the total % Associated query proteins within the main species/clade (under "Species Composition") is equal to the sum of the % of Total Consistent and % Total Inconsistent lineage placements (under "Consistency Assessment").

I have one protein set that gets the following results:

CONSISTENCY ASSESSMENT

Number of proteins in the whole proteome: 46469

#Consistent lineage placements
Total Consistent: 21176 (45.57%)
Consistent, partial hits: 9008 (19.38%)
Consistent, fragmented: 2953 (6.35%)

#Inconsistent lineage placements
Total Inconsistent: 4310 (9.28%)
Inconsistent, partial hits: 2461 (5.30%)
Inconsistent, fragmented: 645 (1.39%)

#Contaminants
Total Contaminants: 3561 (7.66%)
Contaminants, partial hits: 2629 (5.66%)
Contaminants, fragmented: 269 (0.58%)

#Unknown
Total Unknown: 17422 (37.49%)

SPECIES COMPOSITION

##Detected species

#Main species
Clade: Mollusca
Number of associated query proteins: 3989 (8.58%)

#Potential Contaminants

#Potential contaminant Nº1
Clade: Cercopithecinae
Number of associated query proteins: 3652 (7.86%)

I was expecting the associated query proteins to be at 50% or so, and I don't understand where the 8.58% came from, with only 3,989 protein seqs in the Mollusca.fasta file output.

[YanNevers]

Dear @glmaala ,

Thank you for bringing this to my attention. I do not believe this is an error but this points to a lack of clarity in OMArk results. I will consider updating the documentation so this result is easier to interpret.

The discrepancy is due to a difference in the way we handle the Consistent categories and the genes we can unambiguously attribute to a clade.

In the consistency assessments, the category are defined like this:

• Consistent -> the protein is found as part of a gene family (HOG) defined in a clade at the level of Mollusca and above as long it is represented in Mollusca.
• Inconsistent -> the protein is found as part of a gene family that is not represented in Mollusca and associated with other clades, with no clear pattern.
• Contamination -> the protein is found as part of a gene family that is not represented in Mollusca and associated with other clades, with a clear pattern of clade overrepresentation.
• Unknown -> No homolog could be identified

In this case, OMArk detects contamination (although in this particular case, with contamination from Cercopithecinae, it is possibly a mistake from OMArk induced by other bias in the sequences).
OMArk also identifies "Associated proteins": they are the protein than could unambiguously be attributed to each of the detected species/clades. They are the proteins placed in the species tree in clade that share a common ancestor with only one of the detected species/clades. When no contamination is detected, this is all proteins that are not Unknown. When contamination is detected, this is a lower proportion of the dataset.

In this case, only proteins associated to the taxonomic level of Protostome and below would be unambiguously attributed to Mollusca. Indeed, even proteins mapped to gene family that exist in Mollusca but are defined at the level of Bilateria - classed as Consistent - would not be unambiguously attributed to Mollusca because they could instead be attributed to the other bilaterian clade, Cercopithecinae.

As a result, the "Number of associated query protein" will always be lower than the proportion of Consistent genes.

I hope this clears it up.

In your specific results, I would consider the Consistent genes to be the most "trustable" gene models (although they might still be fragmented). I would also take a look at those genes that are detected as contaminant, in my experience erroneous contamination detection may be due to proteins with bias in amino-acid composition or by an abundance of TEs detected as proteins.

Best wishes,
Yannis