Bioinformatics Seminar
Time:
Venue: Na
15 May 2018
NaContext-specific functional analysis of systems in cancer
Dharmesh BhuvaWEHI Bioinformatics
Most current approaches that analyse transcriptional co-expression through network inference focus on a single condition (e.g. disease) ; motivated by the assumption that co-expression networks are context-specific and therefore vary across conditions. I have been investigating the comparative analysis of co-expression networks across conditions (e.g. disease vs normal) ; as differences between these conditions may provide better insight into the functional behaviour of a biological system ; similar in principle to a differential expression analysis. A range of methods have been proposed for the inference and identification of differentially co-expressed genes ; however their applications have been limited as there is no consistent framework for benchmarking and evaluating different methods. Difficulties include incomplete data and missing prior knowledge (i.e. no "perfect" gold-standard network) ; thus I have developed a simulation-based framework to evaluate the performance of 12 inference methods in the context of regulatory networks. I will discuss how previous evaluations may have underestimated the performance of inference methods due to the metrics applied and gold-standard networks used ; and explore some of the network features which are associated with the ability of different inference methods to recapitulate underlying network structure.;;