Curcumin is the yellow pigment of Curcuma longa that irreversibly inhibits the activity of thioredoxin reductase (TrxR) and forms adduct. TrxR, a homodimeric enzyme with E and F chains, is responsible for redox control of the cell and defense against oxidative stress. It is also well known that TrxR is over-expressed in tumor cells. Hence TrxR is a promising target for curcumin based therapy. Binding site of TrxR for curcumin is at the interface of homodimers. In the present study, naturally occurring curcuminoids and forty four synthetic analogs of curcumin were docked with SP/XP glide suite and E-pharmacophore was simulated. E-pharmacophore of both chains has shown three donor features and one acceptor feature. 3D atom based QSAR models have been proposed for the two series of curcumin analogues of known IC50values. The data obtained indicates that the training set model is quite efficient to predict the test set of data. Obtained models and ADMET prediction could be employed for design and synthesis of more potent inhibitors of TrxR.