Increasingly, combinatorial libraries are used to screen large numbers of complex chemically- and physicallypatterned polymers against cell response. There is a strong need to extract meaningful relationships between cell function and material surface features from these experiments. A novel high-throughput cell-material screening strategy, based upon local cell-feature analysis (LCFA) was applied to screen osteoblast proliferation behavior on combinatorial libraries of phase-separated PLGA and PCL. Traditional factor importance analysis, which uses summary statistical inference to identify significant variables, indicated that one controlling material surface feature was PCL diameter. However, the summary statistic analysis was unable to uncover more subtle relationships. The LCFA method, based on histograms of distances between cells and microstructures, was able to identify non-linear, discrete relationships between proliferation, PCL diameter, and cell-PCL distance. LCFA provides an advantage in that a distribution function is not assumed, but rather is developed from the data. Using these results, we propose a model for classifying the material-microstructure interactions, in which small PCL islands far from the cell nucleus act as holders for attachment and large islands close to cells act to shape the cell.