cgMLST for diverse genus

[frschwa]

Hello developers,

thank you for the great tool!

I have been using chewBBACA to create cgMLST schemes for three different species and then validated the schemes by typing a few thousand genomes and calculating the percentage of loci present in those genomes. In both cases >99% of genomes contained >95% of loci, so I consider these schemes stable.

I now wanted to create a genus-wide cgMLST. The genus is quite diverse, so I expected a lower number of loci in the cgMLST95. When I tested the stability of the scheme, I found that 137 of 723 loci are in less than 50% of genomes. For some genomes that can be explained by the fact, that I did not include a genome of that exact species for schema creation. But even for species, for which I included several genomes less than 80% of loci are present.

Have you experienced something similar, when creating schemes for more diverse groups?

Thank you for your help!

[rfm-targa]

Hello @frschwa,

Thank you for your interest!

If the genus is very diverse, you may need to adjust the parameter values used by chewBBACA to identify alleles, such as the BLAST Score Ratio (BSR), whose default value is 0.6. At the genus level, some alleles for a locus may be too divergent to be detected with a BSR value of 0.6. You should change the parameter values to adjust for the greater sequence diversity at the interspecies level. For example, you can change the BSR to 0.55 to allow for the detection of more divergent alleles.

It is also important to choose the best method to define the set of target loci. Did you select a set of genome assemblies from multiple species to create an initial schema with the CreateSchema module? Or did you select loci from the species-level schemas to create a genus-level schema?

The method used to define the target loci and the representative alleles is extremely important and may require a little bit of trial and error to get good representativeness at the genus-level. If sequence divergence between different species is very high, chewBBACA may fail to identify equivalent loci between those species.

The CreateSchema module defines an initial schema, which we call a schema seed, with only one representative allele per locus. chewBBACA uses the representative alleles to identify novel alleles when performing allele calling with the AlleleCall module. The first representative allele for each locus in the schema selected by the CreateSchema module will come from a single genome, a single species. If the alleles in other species' genomes are very different, chewBBACA may not be able to identify them when comparing against the representative allele in the schema. Greater divergence can also cause chewBBACA to define separate schema loci for the same gene identified in each of the species.

In my view, you should explore multiple options to assess what gives the best results. You can 1) select a set of representative/reference genomes that includes all or most species and use that to create a schema seed with the CreateSchema module, 2) select a small set of representative/reference genomes for multiple species that are sufficiently similar for chewBBACA to be able to identify equivalent genes between those species (you can compute the ANI for all pairs of genomes to identify the closest genomes between species), 3) create FASTA files containing representative alleles of each gene for multiple species and use the PrepExternalSchema module to adapt those FASTA files into a schema compatible with chewBBACA. Exploring these options, while also varying parameters such as the BSR, may help in creating a schema that performs well at the genus-level.

Let us know if you have further questions.

Best regards,

Rafael

[frschwa]

Thank you for your detailed answer, @rfm-targa !
I selected 928 assemblies from 49 species and used CreateSchema for this schema. I will try what you recommended and then get back to you.