Simplex-centroid design along with artificial neural network coupled with genetic algorithm (ANN-GA) was applied to maximize exoglucanase production by Penicillium roqueforti ATCC 10110 under solid-state fermentation (SSF), using a blend of agroindustrial wastes as substrate. The first statistical treatment determined the ideal contents of green coconut shell, corn cob, and sugarcane bagasse in the substrate, which were 0.44, 2.06, and 2.50 g, respectively. The optimum conditions by the ANN-GA were obtained as follows: 24 h, 21 °C, and 8.1 and 81.0% for the time, temperature, pH, and moisture, respectively. Moreover, the predicted and the experimental values of exoglucanase activity were 267.94 and 268.58 IU/g, respectively. The optimization process increased the enzyme activity by up to 1263% compared with the preliminary analysis using individual substrates, demonstrating the high efficiency of the algorithms on predicting and optimizing enzyme production. Biochemical characterization demonstrated good thermostability, basic pH stability, halotolerance, and increased enzyme activity in the presence of metal ions (Co²⁺, Ca²⁺, Mg²⁺, and Fe²), solvents (ethanol and dichloromethane), and organic compounds (EDTA, Triton-X, and lactose,). These results indicate the algorithm efficiency for enzyme production purposes.

Graphical abstract

采用单形质心设计以及人工神经网络和遗传算法（ANN-GA），以最大化罗氏青霉产生的外切葡聚糖酶ATCC 10110在固态发酵（SSF）下，以农业工业废料的混合物为底物。第一次统计处理确定了基质中绿色椰子壳，玉米芯和甘蔗渣的理想含量，分别为0.44、2.06和2.50 g。通过ANN-GA获得的最佳条件如下：时间，温度，pH和湿度分别为24 h，21°C和8.1和81.0％。此外，外切葡聚糖酶活性的预测值和实验值分别为267.94和268.58 IU / g。与使用单个底物进行的初步分析相比，优化过程使酶活性提高了高达1263％，证明了算法在预测和优化酶产生方面的高效性。²⁺，Ca 2 ⁺，Mg ²⁺和Fe ²），溶剂（乙醇和二氯甲烷）和有机化合物（EDTA，Triton-X和乳糖）。这些结果表明用于酶生产目的的算法效率。

图形概要