“Estimating effective connectivity in human negative emotional network and inferring the best causal model reveals potentially important differences between subjects with 5-HTTLRP long- and short-allele. We estimate effective connectivity among four brain regions that are activated during the processing of facial affect, e.g., fear: left and right amygdale (L and R, respectively), caudal supragenual and rostral subgenual portions of anterior cingulate cortex (rACC and cACC, respectively). Dynamic causal modeling is used to make inferences about effective connectivity. Bayesian model selection is also used for finding the best model from a set of plausible alternative models. The stimuli paradigm presents fearful facial expressions to a set of 37 subjects who have either two copies of the 5-HTTLRP long-allele or one copy of the 5-HTTLRP short-allele carriers (15 subjects of long-allele and 22 subjects of short-allele). Our results revealed that a fully connected network with no modulated connection was the winning model for long-allele whereas a fully connected network with two modulated connections was the winning model for short-allele individuals. The left to right amygdale (L-R) connections were significantly different between long- and short-allele (p < 0.0194). Modulating factor on L-R was found nonzero (p < .05). In conclusion, this study has shown significant differences between the two groups (long- and short-allele) during processing of negative emotions. Further, it has furnished an important role for modulatory input and left amygdala as discriminatory factors between the two groups.”