Computational theory: statistics, machine learning, modelling and data integration algorithms

Develop the computational theory behind the modelling strategies.

Statistical Machine Learning (SML) is acknowledged as the most important driver of the technology underpinning the success of companies such as Yahoo, Google, Amazon, Microsoft, and Facebook. Furthermore, SML continues to play an essential role in the genomic revolution, being at the heart of the sequencing data analysis algorithms currently employed, as well as being at the core of analytical computational methods applied to all other –omics data (transcriptome, proteome, metabolome). Efficient and robust algorithms make many of these methods effective. We see the future of healthcare, as described in ITFoM, as no different: it will be fundamentally reliant on SML and on efficient and robust algorithms. In addition to SML techniques, we will develop the fundamental approximate inference techniques in large scale mechanistic models that enable their simulation and parameter learning. WP5 is therefore concerned with all aspects of inference requirements for the ITFoM modelling approaches, developing where necessary novel methodologies, both statistical and computational, exploiting the availability of high performance computing.

Evaluation on test cases

Ultimately, the WP5 methodologies need to demonstrate the clinical utility and relevance of the results they produce. To this end, clinical and analytical data sets from appropriate test cases identified in the literature, from the ITFoM reference data set and the ITFoM use cases will be processed using selected methods and models. Medical experts will then evaluate the results yielded and determine their impact in clinical situations, processes and decisions for both patient and physician.