e23142 Background: Anticancer target drugs (ATDs) specifically bind and inhibit molecular targets that play important roles in tumorigenesis. More than 150 different ATDs have been approved for clinical use worldwide, and the clinicians are faced with the problem of choosing the best therapeutic solution for each patient. The problem of efficient ATD selection remains largely unsolved and personalized approaches are needed to select the best ATD candidates for individual patients. Methods: We propose a new approach termed OncoFinder. It is based on digesting gene expression profiles for the analysis of activation of intracellular signalling pathways as a marker for the selection of target therapies. The original bioinformatic algorithms were integrated with the databases featuring molecular drug targets, compositions of signalling pathways, including the functional role of each gene product, for more than 1700 pathways (Buzdin, Front.Genet 2014; Ozerov, Nature Communications 2016). Results: We showed that pathway activation strengths are more stable and reliable biomarkers of cancer than the expressions of individual genes. OncoFinder allows to detect changes at the level of pathway activation and to predict the effectiveness of drugs based on the knowledge of their molecular targets. We applied it to find new biomarkers of clinical response to the ATD cetuximab; for modelling the combined chemotherapy of acute myeloid leukemia and combined anti-VEGF/BRAF therapy of melanoma. For two unrelated datasets obtained for colon cancer patients before treatment with the ATD bevacizumab, we were able to distinguish between those who responded to treatment and not (p < 0.01). We next assayed biopsies for kidney cancer patients with known responses to the ATD sorafenib. The responders and non-responders showed a significant difference (p = 0.02). Finally, the OncoFinder platform was prospectively used for decision making support to patients with advanced metastatic solid tumors (n = 23). The efficiency of the ATD treatment was 61% (complete + partial response, RECIST). Conclusions: OncoFinder method may be effective for predicting response to ATD based on high throughput gene expression profiles.
Simpler Evaluation of Predictions and Signature Stability for Gene Expression Data