Choosing specific candidate genes for phylogenetic tree construction

Question

[Penny Liu]

I want to compare our strains VGH107 with other S. marcescen strains phylogenetic relationship. This selection contained both nonpathogenic (environmental, agricultural, and industrial) as well as pathogenic strains of various pathotypes.

From my understanding, 16S sequences are not suitable to analyze inter-strain relationships within a species or between closely related species. Therefore, I used the Multi-Virulence-Locus Sequence Typing (MVLST) method to discriminate species.

For the phylogenomic analysis, the combination of nine virulence gene fragments (aprA, hlyA, hlyB, hlyIII, luxS, phoP, phoQ, shlA, and shlB) were selected. Subsequently, tree was reconstructed using the Maximum Likelihood (ML) method (MEGA 7) based on Tamura-Nei model with concatenated nucleotide sequences of nine virulence genes.

I expected to divide nonpathogenic and pathogenic group. Apparently, the results show nonpathogenic and pathogenic strains mixed together. How do I effectively separated two group?

Figure. Taxonomic classification of S. marcescens VGH107

Circle color refers to nonpathogenic. Green indicates environmental S. marcescens strains. Blue indicates agricultural strains. The industrial strains are marked pink. Besides, unmarked indicate human pathogen strains.

[Lars Juhl Jensen]

I can think of two things to look into. The first is if the nine virulence genes give congruent results. If you make trees for each of them separately, do the resulting trees look similar or does it look like the different virulence genes may have different evolutionary histories due to lateral gene transfer? The second is if these genes are even the ones that make the difference between virulent and non-virulent strains. If you look at the presence-absence pattern of genes are there any genes that are present in virulent strains but not in the non-virulent ones? Are they by any chance encoded by plasmids or other mobile elements?

[Penny Liu]

Thank you so much for providing me with useful comments. Choosing the nine virulence genes, or even more virulence factors (VFs) are belong to feature selection or clustering problems.

Moreover, the currently selected of nine virulence genes may have degree of discrimination but I think it's not perfect enough. Because some of the pathogenic strains not clustered within cluster I (with VGH107).

For this issue, we intend to reconstruct the trees for each of VFs separately, just like you said, observed different virulence genes may have different evolutionary histories. Finally, choose the discriminative virulence genes.

[Penny Liu]

I'm a little confused by your last comment.

What is the main purpose of finding plasmids or other mobile elements?

Thanks again!

[Lars Juhl Jensen]

What you see is a rather scattered distribution of virulent strains over the phylogenetic tree. A simple explanation for this would be that at least one important virulence factor is carried by a mobile element, for which reason it - and thereby virulence - does not follow vertical descent.

[natasha.sernova]

See this site if you have not seen it before.

http://www.antimicrobe.org/b26.asp

I would study Serretia marcescens drug-resistance and genes responsible for it.

[Penny Liu]

You are right! This study we want to analyze virulence and resistance from Serratia marcescens. (cite)

However, in this post we aim to reconstruct the phylogenetic history of the Serratia marcescens.

Any comments and suggestions would be appreciated :)

[Asaf]

1. Do the non-pathogenic strains have the genes you selected?
2. In the cases of E. coli and Shigella and Bacillus cereus and thuringienesis the pathogenic strains are mixed with the non-pathogetic strains phylogenetically and as suggested above acquired the virulent factors. I think that the phylogenetic tree should be built using house-keeping genes and each virulence factor should have its own tree to see how it spread in the different strains.

[Penny Liu]

non-pathogenic and pathogenic strains are contained all 9 virulence genes.

Answer 1

[kloetzl]

I expected to divide nonpathogenic and pathogenic group. Apparently, the results show nonpathogenic and pathogenic strains mixed together. How do I effectively separated two group?

That sounds like you are looking for a method to support your expectations. Beware, this is not the scientific method.

Also, why limit yourself to specific genes at all? That will introduce some kind of selection bias. Just use the whole genome instead.

[Penny Liu]

The reason is multiple sequence alignment (MSA) program - ClustalW is extremely limited when in regards to multiple whole genome sequencing.

[kloetzl]

A tool should not limit your research. Use another tool! I suggest mugsy for bacterial whole genome alignments.