MLST - output of genotypes with several alleles detected.

[evezeyl]

When 2 or more alleles are detected at the same gene, whether those alleles are new or existing, this should appear in the results.
As per today it doesnt. Moreover in this case the CC group must remain undertermined, which is not the case until now

• This will allow to Detect where things that are weirds, incl. possible intraspecific contamination
• PS: might be worth to update the database for MLST also (seems that the version has been updated)
  You can use as example Listeria . VI56247 it should have 2 alleles detected on dapE gene

PS: I will recheck the scheme I used to be 100 sure that things are working as they should also in ALPPACA .... so we need to combine efforts

[evezeyl]

Some additional explanations:
Old version of MLST software use "lmonocytogenes" scheme (BIGSDB - Pasteur)
While the new version of MLST software use "listeria_2" scheme - apparently a change of definition of dapE - we need to trace back. The listeria_2 scheme give rise to the multiple alleles during typing while the lmonocytogenes scheme do not.

Asked the question on Github regarding the orgin of the listeria_2 scheme in MLST repo: <tseemann/mlst#128 question posed about origin of this scheme>

[evezeyl]

So I made an admixed sample from VI55314 and VI55689 reads.
Launching pipeline MLST to see if detected correctly
Also launching the quality QC assembly and QC reads to see if we detect admixture or not, to be sure

[evezeyl]

So first tests with confindr did not detect admixture (some samples did not run yet)

• find out if those that fail are the ones with weird alleles? (no contamination detection, failure test because memory likely on pc)

• Admixture detection tests need to be relaunched when the sever is up again (OEIO mixed sample for testing)

Conclusion for now:

• nothing to change to the tool, no update. :)
• opened an issue for ALPPACA cgMLST track as it uses the latest version of MLST - DONE

[evezeyl]

Results on Admixted isolate (different CCs from VIGAS-P) - No contamination detected by confindr on the sample used -

• VIGAS MLST pipeline (version 2.16.1 usinglmonocytogenes MLST scheme).

• For running MLST I used two assemblies from different VIGAS-P pipeline and tested with MLST software - 2 different versions- 1. using the same sheme: lmonocytogenes that was used in VIGAS-P MLST pipeline AND with the Listeria 2 (sheme that gives multiple alleles for dapE and made me doubt about contamination for some isolates - scheme that is used in ALPPACA cgMLST track).

• Note here I chose isolates where no ambiguous call was made for dapE to create the admixed sample. The admixed sample was created by concatenating the reads from VI55314 and VI55689. I did test if I could detect intra-specific contamination using confindr rMLST pipeline, and could not detect any (only one SNP position - possible variant), "contamination" was detected in the admixed sample.

• to run MLST on SAGA - I used two assemblies created by different assembly pipelines from VIGAS-P, with assembly annotation pipeline and contig assembly QC.

NB: lmonocytogenes scheme corresponds to the scheme a I used previously and that I know corresponds to the Scheme that was on BIGSDB pasteur at least until 2021).

Method	FILE	SCHEME	ST	abcZ	bglA	cat	dapE	dat	ldh	lhkA
MLST 2.16.1	ADMIXED-VIGAS-P MLST pipeline	lmonocytogenes	-	6	6?	6	4	1	4	1
MLST mlst 2.19.0	ADMIXED-contigs_assembly_annotation.fasta	lmonocytogenes	-	6	6?	6	5	1	4	1
MLST mlst 2.19.0	ADMIXED-contigs_assembly_QC.fasta	lmonocytogenes	-	6	6?	6	4	1	4	1
MLST 2.23.0	ADMIXED-contigs_assembly_annotation.fasta	listeria_2	-	6	6?	6	5	1	4	1
MLST 2.23.0	ADMIXED-contigs_assembly_QC.fasta	listeria_2	-	6	6?	6	4	1	4	1
MLST 2.16.1	VIGAS-P MLST pipeline VI55314	lmonocytogenes	9	6	5	6	4	1	4	1
MLST mlst 2.19.0	VI55314.fasta	lmonocytogenes	9	6	5	6	4	1	4	1
MLST 2.23.0	VI55314.fasta	listeria_2	9	6	5	6	4	1	4	1
MLST 2.16.1	VIGAS-P MLST pipeline VI55689	lmonocytogenes	121	7	6	8	8	6	37	1
MLST mlst 2.19.0	VI55689.fasta	lmonocytogenes	121	7	6	8	8	6	37	1
MLST 2.23.0	VI55689.fasta	listeria_2	121	7	6	8	8	6	37	1

[evezeyl]

Conclusions and consequences:

• good: CC is not determined in any case for the admixed sample
• good: assemblies by 2 different methods - provided same typing for the "pure samples"
• bad: different assemblies of admixed sample result in different typing, possibility deterministic downsampling of the reads (as I concatenated in order, might only have downsampled from the second sample and not from the first // or one sample predominates at assembly), and multiple alleles are not detected in any case ....
• bad: movel alles discovered by MLST are not added clearly - but it is indicated in the detail novel allele file that novel allele was discovered for dapE, it is just not reported

Action:
Suggest that we reopen the QC issue that I had opened conserning intra-genus contamination where I mentionned inplementing confindr (or any means to detect potential intraspecific contamination) as this could create problems ... and remain undetected.

Sugest finding a way to output that novel alleles have been detected but not reported using MLST

Discussion?

• What scheme should we use ? Listeria 2 appears to report the multiple alleles detected (dapE), I think it did not for the admixed sample probably because assembly eliminated part of the problem, and its consistent otherwise with the lmonocytogenes. The resulting would be however many more ST would be unknown ... (or keep it as it - and base our results more on cgMLST) - Does it require a comparison with benchmarking isolates from which CC is known from the lab ?