Background: Genetic factors have been suggested to affect the efficacy of working memory training. However, few studies have attempted to identify the relevant genes. Methods: In this study, we first performed a randomized controlled trial (RCT) to identify brain regions that were specifically affected by working memory training. Sixty undergraduate students were randomly assigned to either the adaptive training group (N = 30) or the active control group (N = 30). Both groups were trained for 20 sessions during 4 weeks and received fMRI scans before and after the training. Afterward, we combined the data from the 30 participants in the RCT study who received adaptive training with data from 71 additional participants who also received the same adaptive training but were not part of the RCT study (total N = 101) to test the contribution of the COMT Val158/108Met polymorphism to the interindividual difference in the training effect within the identified brain regions. Results: In the RCT study, we found that the adaptive training significantly decreased brain activation in the left prefrontal cortex (TFCE-FWE corrected p =.030). In the genetic study, we found that compared with the Val allele homozygotes, the Met allele carriers' brain activation decreased more after the training at the left prefrontal cortex (TFCE-FWE corrected p =.025). Conclusions: This study provided evidence for the neural effect of a visual–spatial span training and suggested that genetic factors such as the COMT Val158/108Met polymorphism may have to be considered in future studies of such training.