A colleague of mine has performed experimental evolution of some bacteria populations in the presence of several antimicrobial agents. The experiment has been replicated twice (2 blocks containing each 3 replicate populations per condition). We have re-sequenced some individual colonies from these populations at the end of the experimental evolution.
There were 5 genes which were over-represented as having SNPs. In some colonies, there were several of these 5 genes which showed SNPs. In the end, my colleague measured the difference in minimum inhibitory concentration reached at the end of the evolution versus what it was at the beginning (variable "MIC fold-change"). The sample size is not so high in the end, which is typical of this kind of experiment, but I would still like to be able to analyze whether some combinations of mutations were selected in some regimes (more for my own improvement than for the result honestly).
My dataset contains one column per each of these 5 genes with “0” meaning no SNP, “1” meaning SNP. I also added one column which is a concatenation of the affected genes, i.e. ~ the mutations combination. Some combinations emerge only once of course...
I attach a subset of the dataset :
| ID | selection regime | Block | MIC_log2_fold_change | MIC_fold_change_lin | LasR | PhoQ | WbpA | mutations |
|---|---|---|---|---|---|---|---|---|
| Melittin1.1 | melittin | 1 | 6 | 64 | 0 | 1 | 1 | PhoQ-WbpA |
| Melittin1.2 | melittin | 2 | 5 | 32 | 1 | 1 | 0 | LasR-PhoQ: |
| CecropinP11.1 | cecropin | 1 | 1 | 2 | 0 | 1 | 0 | PhoQ |
| CecropinP12.2 | cecropin | 2 | 1 | 2 | 0 | 0 | 1 | WbpA |
| SLM12.1 | SLM | 1 | 2 | 4 | 1 | 1 | 1 | LasR-PhoQ-WbpA |
| SLM12.3 | SLM | 2 | 4 | 16 | 0 | 0 | 0 | 0 |
I have performed a conditional logistic regression using the clogit command from the survival package with each of the 5 genes in turn as a response variable (presence/absence), and the other remaining genes as explanatory variables (presence/absence), with the block as a random factor.
I would not be surprised if there was a much better alternative, which is why I am asking the opinion of the community. I would really love to be able to ask for mutations combination ~ selection regime +(1|block) for example.
Ultimately it would be great to know whether MIC fold change is affected by the mutation combination or individual mutations !
