4-Hydroxyphenylpyruvate dioxygenase (EC 1.13.11.27, HPPD, a Fe(II)/alpha-ketoglutarate dependent oxygenases), is a popular herbicide target. In this work, two pharmacophore models based on common molecular charac-teristics (HipHop) and receptor-ligand complex (CBP) were generated for virtual screening for HPPD inhibitors. About 1,000,000 molecules containing diketone structure from PubChem were filtered by Lipinski's rules to build a 3D database. Then the database was screened through combining HipHop model, CBP model, ADMET (ab-sorption, distribution, metabolism, excretion and toxicity) prediction and molecular docking. Subsequently, based on the specific binding mode and affinity of HPPD inhibitors, 4 molecules with high -CDOCKER energy, good aqueous solubility and human safety predicative properties values were screened. From the screening re-sults and combined with previous work, three novel HPPD inhibitors were designed and synthesized through fragment splicing and bioisosterism strategies. Compound IV-a exhibited similar inhibition of Arabidopsis thali-ana HPPD (AtHPPD) and herbicidal activity as mesotrione. Crop selectivity showed that compound IV-a had better crop safety than mesotrione. Comparing the molecular properties, ADMET and molecular docking studies indicated that compounds IV-a exhibited better properties than mesotrione, which could be further modified as novel HPPD inhibitor herbicides.