Carlos Riveros

Carlos Riveros long term research focus is in data mining and knowledge discovery problems in large, diverse and complex data. A particular area of interest is the application of modern massively parallel compute paradigms, such as distributed grid computing, cloud computing/map-reduce, General Purpose Graphical Processing Units (GPUs) to accelerate computation in large and very large biological datasets. Design, implementation, testing and improvement of tools making use of these platforms open new avenues to test biological hypothesis at the individual level, enable more accurate identification of markers of disease, and better characterisation of different disease subgroups with different responses to treatment and survival outcome prediction. Carlos' expertise comes from a diverse career in applied High-Performance Computing for research, industry and academia sectors.

His recent work connects techniques for approximate solutions, generalised data integration models and information and graph theory approaches. It involves the use of evolutionary computation heuristics (genetic and memetic algorithms) applied to obtain approximate solutions, as exact solution methods (for example linear and mixed integer programming optimisation models) become quickly infeasible for larger problem instances.

A related area is the formulation of generalised models for data aggregation and meta-analysis, with mathematical guaranties of separability, to allow for robust identification of biomarkers without loosing internal substructure in the groups being differentiated.

Finally, these techniques are complemented with well-founded graph- and information-theory methods for unsupervised clustering, under different notions of element similarity.

Recent publications include application of these novel methods to the analysis of transcriptomics and genomics data in complex diseases such as Age-Related Macular Degeneration, Alzheimer's Disease, Multiple Sclerosis, Breast and Prostate Cancer.