Optimization of pH and nitrogen for enhanced hydrogen production by Synechocystis sp. PCC 6803 via statistical and machine learning methods

The nitrogen (N) concentration and pH of culture media were optimized for increased fermentative hydrogen (H2) production from the cyanobacterium, Synechocystis sp. PCC 6803. The optimization was conducted using two procedures, response surface methodology (RSM), which is commonly used, and a memory‐based machine learning algorithm, Q2, which has not been used previously in biotechnology applications. Both RSM and Q2 were successful in predicting optimum conditions that yielded higher H2 than the media reported by Burrows et al., Int J Hydrogen Energy. 2008;33:6092–6099 optimized for N, S, and C (called EHB‐1 media hereafter), which itself yielded almost 150 times more H2 than Synechocystis sp. PCC 6803 grown on sulfer‐free BG‐11 media. RSM predicted an optimum N concentration of 0.63 mM and pH of 7.77, which yielded 1.70 times more H2 than EHB‐1 media when normalized to chlorophyll concentration (0.68 ± 0.43 μmol H2 mg Chl−1 h−1) and 1.35 times more when normalized to optical density (1.62 ± 0.09 nmol H2 OD730−1 h−1). Q2 predicted an optimum of 0.36 mM N and pH of 7.88, which yielded 1.94 and 1.27 times more H2 than EHB‐1 media when normalized to chlorophyll concentration (0.77 ± 0.44 μmol H2 mg Chl−1 h−1) and optical density (1.53 ± 0.07 nmol H2 OD730−1 h−1), respectively. Both optimization methods have unique benefits and drawbacks that are identified and discussed in this study. © 2009 American Institute of Chemical Engineers Biotechnol. Prog., 2009