Research Study Group
"Using Multi-omic Cancer Data to Find Ways to Improve the Treatment of Bladder Cancer"
In recent literature on bladder cancer there is call time and time again for better identification of cancer subtypes. Broadly, bladder cancer can be split into two types: muscle invasive and non-muscle invasive. By successfully identifying sub-phenotypes of bladder cancer using multi-omic data, we hope to be better able to personalise cancer care and choose appropriate treatments. To achieve this goal, we analysed the rich data source available to us via The Cancer Genome Atlas, in particular focussing on Methylation, Gene Expression and Copy Number Variation data. We found evidence supporting current research on using the gene AQP1 as a bladder cancer biomarker. We classified a subgroup of patients with a higher mean age, who had highly methylated genes but better survival prognosis. Further, we identified ways in which multi-omic data could be used to provide more detail than the currently used TNM system. Finally, we performed Multiple Dataset Integration (MDI) to provide a way of combining the different data types.
"Understanding Medellín’s Education System through Data Analysis and Modelling"
Medellín is the second city of Colombia. Once described by TIME as the most dangerous city in the world, it has become a world model of social inclusion, and a nucleus, for science and technology in Colombia. Medellín has funded various social strategies with education and culture as priorities that guide them. One of the goals that Medellín has is to improve the quality of teaching to redesign the education system and the role of teachers. In order to do that, a variety of quantitative and qualitative data has been collated, comprising of teacher demographics, quality of education and student performance. In the present work some exploratory data analysis has been performed as a tool to produce useful and evocative pictures that can help the polycimakers f Medellín to understand what is going on in the city and how policies can be improved. Furthermore the relationships among the variables and their dynamics has been described by means of a bayesian network, and used to predict the possible results of policies.