Recently, artificial intelligence and more specifically machine learning has been granted an ever increasing attention for their ability to solve diverse challenging tasks in various fields. Especially, their use in medicine and genomics has been proven as very beneficial for addressing clinical relevant questions and moving towards a patient tailored care. For this study, we relied on the publicly available TCGA data and used RNA-sequencing data from 4615 tumor biopsy sample from 10 different tumor types. To achieve the generation of a meaningful gene signature we compared both commonly used and state-of-the-art clustering algorithms that we compared using a wide range of both biological and mathematical assessment scores. The gene signature produced by our proposed approach reports at least 10 times better statistical significance and 35% better biological significance than the ones produced by 5 referential unsupervised clustering methods. Moreover, our experiments demonstrate that our low dimensional biomarker (27 genes) surpass significantly existing state of the art methods both in terms of qualitative and quantitative assessment while providing better associations to tumor types than methods widely used in the literature that rely on several omics data.
E. Battistella, M. Vakalopoulou, T. Estienne, M. Lerousseau, R. Sun, C. Robert, N. Paragios, and E. Deutsch, “Gene expression high-dimensional clustering towards a novel, robust, clinically relevant and highly compact cancer signature,” in IWBBIO 2019, Granada, Spain, May 2019