Microarray data can provide a wealth of knowledge related to predicting a biological outcome. Current research in this area often aims to include information from biological pathways, which can describe gene relationships and interactions, in order to improve understanding of the underlying biological process. Incorporating pathway information is an increasingly attractive option as more knowledge about pathway structure and function becomes available online in publicly accessible databases. Here, a novel method for identifying important biological pathways and genes for classification problems is proposed using guided regularized random forests (GRRF). This approach allows for ranking both pathways and genes, and the regularization of GRRF can help researchers select a compact set of features related to the biological outcome. The proposed methodology is able to successfully identify significant biological pathways and genes and achieve low misclassification rates in both simulation experiments and the analysis of a breast cancer dataset.