FMS-like tyrosine kinase 3 (FLT-3) is strongly correlated with acute myeloid leukemia, but no FLT-3-inhibitor cocomplex structure is available to assist the design of therapeutic inhibitors. Hence, we propose a dual-layer 3D-QSAR model for FLT-3 that integrates the pharmacophore, CoMFA, and CoMSIA. We then coupled the model with the fragment-based design strategy to identify novel FLT-3 inhibitors. In the first layer, the previously established model, Hypo02, was evaluated in terms of its correlation coefficient (r), RMS, cost difference, and configuration cost, with values of 0.930, 1.24, 106.45, and 16.44, respectively. Moreover, Fischer's cross-validation test of data generated by Hypo02 yielded a 98% confidence level, and the validation of the testing set yielded a best r value of 0.87. The features of Hypo02 were separated into two parts and then used to screen the MiniMaybridge fragment compound database. Nine novel FLT-3 inhibitors were generated in this layer. In the second layer, Hypo02 was subjected to an alignment rule to generate CoMFA- and CoMSIA-based models, for which the partial least-squares validation method was utilized. The values of q(2), r(2), and predictive r(2) were 0.58, 0.98, and 0.76, respectively, derived from the CoMFA model with steric and electrostatic fields. The CoMSIA model with five different fields yielded values of 0.54, 0.97, and 0.76 for q(2), r(2), and predictive r(2), respectively. The CoMFA and CoMSIA models were used to constrain 3D structures of the nine novel FLT-3 inhibitors. This dual-layer 3D-QSAR model constitutes a valuable tool to easily and quickly screen and optimize novel potential FLT-3 inhibitors for the treatment of acute myeloid leukemia.

fmslike tyrosin kinas flt strong correl acut myeloid leukemia fltinhibitor cocomplex structur avail assist design therapeut inhibitor henc propos duallay dqsar model flt integr pharmacophor comfa comsia coupl model fragmentbas design strategi identifi novel flt inhibitor first layer previous establish model hypo evalu term correl coeffici r rms cost differ configur cost valu respect moreov fischer crossvalid test data generat hypo yield confid level valid test set yield best r valu featur hypo separ two part use screen minimaybridg fragment compound databas nine novel flt inhibitor generat layer second layer hypo subject align rule generat comfa comsiabas model partial leastsquar valid method util valu q r predict r respect deriv comfa model steric electrostat field comsia model five differ field yield valu q r predict r respect comfa comsia model use constrain d structur nine novel flt inhibitor duallay dqsar model constitut valuabl tool easili quick screen optim novel potenti flt inhibitor treatment acut myeloid leukemia