The aim of our study was to illuminate the potential role of brain-derived neurotrophic factor (BDNF) in autism spectrum disorder (ASD). We measured the circulating levels of BDNF in serum and BDNF gene (Val66Met) polymorphisms, in which two indicators were then compared between ASD and normal controls. A total of 82 drug-naïve ASD children and 82 age- and gender-matched normal controls were enrolled in the study. Their serum BDNF levels were detected by the ELISA. BDNF Val66Met polymorphism genotyping was conducted as according to the laboratory's standard protocol in laboratory. The ASD severity assessment was mainly determined by the score of the Childhood Autism Rating Scale (CARS). ELISA assay showed that the mean serum BDNF level of children with ASD was significantly (P < 0.0001) higher than that of the control cases (17.75 ± 5.43 vs. 11.49 ± 2.85 ng/ml; t = 9.236). Besides, the serum BDNF levels and CARS scores (P < 0.0001) were positively related. And, the BDNF genotyping results showed that there was no difference between the ASD cases and the control. Among the children with ASD, the mean serum BDNF level of Met/Met group was lower than other groups. According to the ROC curve generated from our clinical data, the optimal cutoff value of serum BDNF levels, an indicator for diagnosis of ASD, was projected to be 12.50 ng/ml. Thus, it yielded a corresponding sensitivity of 81.7 % and the specificity of 66.9 %. Accordingly, area value under the curve was 0.836 (95 % CI, 0.774-0.897); the positive predictive value (PPV) and the negative predictive value (NPV) were 70.1 and 79.1 %, respectively. These results suggested that rather than Val66Met polymorphism, BDNF was more possible to impact the pathogenesis of ASD.