Incomplete factorial and response surface methods in experimental design: Yield optimization of tRNA^TrP from in vitro T7 RNA polymerase transcription

We have studied the yield of Escherichia coli tRNA^trp obtained from in vitro T7 RNA polymerase transcription using incomplete factorial and response surface methods. Incomplete factorial experiments were first used to estimate the relative impact of six variables on the yield of tRNA^TrP. Fifteen trials were performed according to a balanced and randomized design. The correlation between observed yield and all experimental variables was identified by stepwise multiple linear regression analysis. The concentrations of T7 RNA polymerase, DNA template, NTP and MgCl₂ proved to be significantly correlated with the yield of tRNA_Trp. We then optimized the yield with respect to each of these four variables simultaneously with a designed, response surface experiment based on the Hardin-Sloane minimum prediction variance algorithm. Twenty experiments were performed, in duplicate, to sample the quadratic surface relating the yield to the four significant variables. Coefficients of the quadratic function with all two-factor interactions were evaluated by stepwise regression using least squares, and significant coefficients were retained. Partial differentiation of the resulting quadratic model showed it to possess an optimum. Transcription performed at the corresponding conditions yielded 6-fold more tRNA^Trp than the initial conditions, confirming the predictive value of the experimentally determined response surface.