Characterization of OncoScore™ Modules in Cancer Cell Lines

The OncoScore™ approach can be applied not just to clinical sample cohorts, but also to pre-clinical model systems, including animal models and cell lines, to provide a clinically relevant biological classification, such that in vitro treatment response data can be correlated with biological subtypes of breast cancer. This approach is useful when existing microarray data is available for a collection of model systems. Module scores are computed for a set of model systems using Compendia’s module mapping approach that seeks to intersect co-expression modules defined from clinical tumors with co-expression patterns observed in the model system dataset. After modules are linked to model system co-expression patterns, intersecting genes are identified and relative module scores are computed. In addition to quantitative module scores, a K-means clustering approach is used to classify cell lines as positive or negative for each mapped module. Please note that some modules will inherently not map to cell line model systems, as they correspond to biological features of the tumor not present in cell culture (e.g. extracellular matrix gene expression programs).