P04.72 Using Mendelian randomization to find potential novel drug targets for the treatment of glioma. (19th September 2018)

• Record Type: Journal Article
• Authors: Robinson, J W; Zheng, J; Tsavachidis, S; Haycock, P; Bondy, M; Relton, C; Martin, R; Smtih, G D; Kurian, K M
• Is Part Of: Neuro-oncology. Volume 20(2018)Supplement 3
• Page Start: iii296
• Page End: iii297
• Publication Date: 2018-09-19
• Subjects: Brain Neoplasms -- Periodicals; Brain -- Tumors -- Periodicals; Brain -- Cancer -- Periodicals; Nervous system -- Cancer -- Periodicals; 616.99481
• DOI: 10.1093/neuonc/noy139.306
• Languages: English
• ISSNs: 1522-8517

Abstract

Abstract: Background: There are about 12, 000 new brain tumours each year in the UK. Around 50% of these are gliomas, of which the majority are the highly aggressive glioblastoma multiforme (GBM) subtype. The median survival rate of GBM is 15 months. The treatment regime for glioma has remained relatively static since the introduction of temozolomide; the median survival rate for patients has increased relatively little since, despite advances in potentially helpful fields such as genomics and bioinformatics. Current methods of drug discovery are unsustainable and inefficient. Costs for development of a new drug can run up to around US$802 million per drug, taking around 12 years before approval for marketing. It is not always guaranteed a drug will make it to market; many fail in stage III trials due to insufficient safety or efficacy. A paradigm shift in the industry is necessary for innovation to remain high. We propose the use of Mendelian randomization (MR) - a statistical method that uses genetic anchors to infer causal relationships between exposure (e.g. phenotype) and outcome (e.g. disease) - as a potential tool to find novel drug targets for glioma. MR produces more reliable results compared to standard observational epidemiological studies, as MR is less susceptible to many bias, e.g. reverse causation and confounding. Specifically, we propose that results from MR analyses may prioritise which drug targets are most likely to pass efficacy tests. Material and Methods: We used protein quantitative trait locus (pQTL) data on plasma proteins as genetic instruments and glioma data from the recent genome wide association study as the outcome (12, 497 cases, 18, 190 controls from six studies). To increase reliability of the instruments, we focused on 2, 346 significant pQTLs (Bonferroni-corrected P threshold of 5x10 -8 ) with no heterogeneity across studies and performed two-sample MR. For MR results that reached statistical significance, we further evaluated the findings by performing Steiger filtering to assess the directionality of effects (e.g. whether the proximal pQTL effect is on the protein which then influences glioma, or vice versa). All analyses were performed in the TwoSampleMR R package. Results: The MR analysis identified three proteins strongly associated (1, 398 tests, P threshold of 3.58x10 -5 ) with glioma: CD36, SEMA6A and CDH5. The top hit, CD36, is known to play a part in angiogenesis and its expression has been found to affect prognosis for patients with GBM. These associations are necessary but not sufficient to prove causality, as potential horizontal pleiotropy remains an alternative explanation. Conclusion: The results provide three potential new targets for the prevention of GBM progression, provided the causes of incidence are the same as progression. Upon further research, this could be evidence that MR may be a useful tool in the drug discovery process and have translational uses.