Estimating the conjugative transfer rate of antibiotic resistance genes: Effect of model structural errors

The spread of antibiotic resistance genes (ARG) occurs widely through plasmid transfer majorly facilitated via bacterial conjugation. To assess the spread of these mobile ARG, it is necessary to develop appropriate tools to estimate plasmid transfer rates under different environmental conditions. Process-based models are widely used for the estimation of plasmid transfer rate constants. Empirical studies have repeatedly highlighted the importance of subtle processes like delayed growth, the maturation of transconjugants, the physiological cost of plasmid carriage, and the dependence of conjugation on the cultures growth stage. However, models used for estimating the transfer rates typically neglect them. We conducted virtual mating experiments to quantify the impact of these four typical structural model deficits on the estimated plasmid transfer rate constants. We found that under all conditions, the plasmid cost and the lag phase in growth must be taken into account to obtain unbiased estimates of plasmid transfer rate constants. We observed a tendency towards the underestimation of plasmid transfer rate constants when structurally deficient models were fitted to virtual mating data. This holds for all the structural deficits and mating conditions tested in our study. Our findings might explain an important component of the negative bias in model predictions known as the plasmid paradox. We also discuss other structural deficits that could lead to an overestimation of plasmid transfer rate constants and we demonstrate the impact of ill-fitted parameters on model predictions.