As the final step of a study aiming at the optimization of culture conditions for the production of carotenoids by red yeasts, a statistically-based experimental design has been applied to assess the influence of selected trace elements on carotenogenesis in Rhodotorula graminis DBVPG 7021. In particular, a central composite design scheme has been used to evaluate the influence of Fe3+, Co2+, Mn2+, Al2+ and Zn2+ (within the range 0-50 ppm) on various responses, namely biomass (B), total carotenoid production (TC) and percentage of specific carotenoids (beta-CAR, beta-carotene; gamma-CAR, gamma-carotene; TN, torulene; TD, torularhodin) on total carotenoids. Second-order polynomial models were calculated and reduced equations were designed by neglecting non-significant (P < 0.01) regression coefficients. Reduced equations were used to calculate the optimal concentration of trace elements in view of maximising the level of B, TC, P-CAR, gamma-CAR, TN and TD. After optimization, average final values total carotenoids (TC = 803.2 mu g/g DW) was about 370% of value observed as central point of the central composite design scheme. Under the same condition, average final values of other responses were: B = 5.40 g/L; beta-CAR = 50.3%; gamma-CAR = 15.4%; TN = 22.7%; TD = 11.6%. All above experimental data are in good agreement with calculated ones, thus confirming the reliability of the proposed empirical model in describing carotenoid production by R. graminis as a function of trace element concentrations. (c) 2005 Elsevier Inc. All rights reserved.