Patient Stratification Case Study: RA
In many instances, a single mechanistic biomarker is not adequate to predict patient response to therapy. Unlike the breast cancer example where the expression of the ER receptor on breast cancer tumors can be used as a clinical diagnostic for treatment with tamoxifen, patients suffering from rheumatoid arthritis do not benefit from a validated diagnostic test to determine if they will respond to anti-TNF therapies.
This case study is based on gene expression data (GSE21537) collected from synovial biopsy tissue of rheumatoid arthritis patients treated with an anti-TNF therapy. The Selventa Discovery Platform (SDP) was used to predict drug response for individual patients using baseline patient gene expression data. Stratification was based on the hypothesis that patients with higher levels of TNF activity would have disease processes more potently driven by TNF, and thus be better candidates for anti-TNF therapy (see poster). TNF activity was inferred from the TNF molecular footprintdescribed in the Selventa knowledgebase, which is comprised of ~1500 genes. The TNF molecular footprint was also refined to a subset of the TNF regulated genes that are altered in RA synovium. Stratification of patients by the inferred TNF activation strength score was effective in separating anti-TNF responding patients from non-responding patients (p<0.05) and the refined molecular footprint improved the separation further (p = 0.001). Inferred TNF activity strength scores correctly predicted responsive patients better than use of the expression of TNF mRNA alone, which did not have significant predictive power (p = 0.169), (Figure 1).