Identification of structural components associated with cytostatic activity in MCF-7 but not in MDA-MB-231 cells

The National Cancer Institute's Developmental Therapeutics Program maintains the screening results obtained in 60 standardized cancer cell lines and contained 37,836 compounds for this study. This dataset has shown to be an outstanding resource for the development of structure-activity relationship (SAR) models describing anticancer activity. We report here a novel SAR modeling approach based on a subtractive protocol to develop models that describe cell type-specific molecular descriptors of cytotoxicity. The goal of this approach is to separate features associated with antiproliferative activity to many cell lines from those that effect only a specific cell type. To assess this approach, we developed SAR models for cytostatic activity against the human breast cancer cell lines MCF-7 and MDA-MB-231 and one differential activity model for compounds that were potent cytostatic agents in MCF-7 cells but relatively inactive against MDA-MB-231 cells. The models were between 72 and 84% accurate when challenged with compounds not in the learning sets. Structural features associated with the differential activity model highlighted how the use of this approach can selectively identify chemical moieties associated with potent cytostatic action to MCF-7 but not to MDA-MB-231 cells. We surmise that outgrowth of this method can facilitate the development of SAR models with sufficient resolution and clarity to identify chemical moieties associated with antiproliferative activity to selective individual cancer types while being innocuous to other cell types.