Paecilomyces tenuipes, one of the commonly used Chinese medicinal fungi, possesses various biological activities. However, the yield of active constituents in P. tenuipes is low. Therefore, it is urgent to screen high-production strain. Protoplast is an important material which has been widely used in filamentous fungi breeding. In this study, a sequential statistical approach was applied to optimize the protoplast preparation conditions from P. tenuipes RCEF4339. Based on single-factor optimization strategy, the optimal condition was obtained as hydrolysis by 2 mg/mL lyticase at 32 ℃ and pH 7 for 3 h, wherein using 0.9 mol/L KCl was used for stabilization of osmotic pressure. Then, key factors (enzymatic hydrolysis time, enzymatic hydrolysis temperature and pH ) were identified by Plackett-Burman design. Finally, the suitable values of three key factors were further optimized by Box-Behnken design. A multiple quadratic regression model was established for fitting the results of Box-Behnken design experiment. Response surface methodology was employed to investigate the mutual effects between candidate factors and response value. Artificial neural network-genetic algorithm were further used to model and optimize the experimental results obtained from Box-Behnken design experiment. The optimum conditions of protoplast preparation of P. tenuipes were hydrolysis by 2 mg/mL lyticase at 28 ℃，pH 6.0 for 3.7 h, wherein 0.9 mol/L KCl was used for stabilization of osmotic pressure. The mean amount of protoplast reached 4.398 2×107 /mL. The rate of protoplast regeneration was 31.25%. The experimental results showed that this method was reliable.