Originally, the Taguchi methods were formulated for the optimisation of industrial processes, where several factors (3 to 50) of complex multifactorial experiments were tested at different levels (Taguchi, 1986). The Taguchi methods use orthogonal arrays to organise the ‘control’ parameters/factors affecting a process and the levels at which they should vary. A particular algorithm (quadratic loss function) is then applied in order to predict the optimum conditions of a process, whilst accounting for performance variations due to ‘noise’ factors beyond the control of the design. In a normal factorial strategy, every parameter should be individually tested at several levels, thus becoming extremely time-consuming, labour-intensive and expensive. The Taguchi methodology allows for testing only a few combinations, therefore dramatically decreasing the total number of experiments and simultaneously identifying the optimum condition of several factors.
Because some functional genes are present only in small fractions of microbial communities, and only few copies can be present in each genome, their detection by classical PCR methods can be challenging. Optimisation of the experimental conditions of a PCR includes the different components of the reaction mix (concentrations of salt, primers, enzyme, DNA template, etc.) as well as the cycling features (time and temperature of the denaturation, annealing and extension steps, number of cycles, etc.). We used this approach for the optimisation of the detection by PCR of functional genes of non-cultivable microorganisms present in environmental samples. In particular, we tested the different parameters involved in a (touchdown/nested) PCR and estimated the optimum settings for the detection of the functional gene pmoA, coding for the putative active site of the particulate methane monooxygenase, involved in the oxidation of methane by methanotrophic bacteria. The application of the Taguchi method allowed the suppression of a nesting step and thus a significant reduction in the amplification time, as well as reagent cost.