Engineered sequence-specific zinc finger nucleases (ZFNs) make the highly efficient modification of eukaryotic genomes possible. However, most current strategies for developing zinc finger nucleases with customized sequence specificities require the construction of numerous tandem arrays of zinc finger proteins (ZFPs), and subsequent largescale in vitro validation of their DNA binding affinities and specificities via bacterial selection. The labor and expertise required in this complex process limits the broad adoption of ZFN technology. An effective computational assisted design strategy will lower the complexity of the production of a pair of functional ZFNs. Here we used the FoldX force field to build 3D models of 420 ZFP-DNA complexes based on zinc finger arrays developed by the Zinc Finger Consortium using OPEN (oligomerized pool engineering). Using nonlinear and linear regression analysis, we found that the calculated protein-DNA binding energy in a modeled ZFP-DNA complex strongly correlates to the failure rate of the zinc finger array to show significant ZFN activity in human cells. In our models, less than 5% of the three-finger arrays with calculated protein-DNA binding energies lower than -13.132 kcal mol(-1) fail to form active ZFNs in human cells. By contrast, for arrays with calculated protein-DNA binding energies higher than -5 kcal mol(-1), as many as 40% lacked ZFN activity in human cells. Therefore, we suggest that the FoldX force field can be useful in reducing the failure rate and increasing efficiency in the design of ZFNs.