Statistical learning for accurate clinical outcomes

Precision medicine is the customisation and tailoring of medical interventions to the individual and has the potential to have an enormous impact on healthcare. Precision medicine offers a path to helping people stay healthy for longer and recover from illness faster. This potential can only be fully realised if we are able to accurately predict the outcome of patients in different clinical scenarios. However the success of this approach relies upon an accurate outcome definition, and the selection of appropriate methods and key features that can classify individuals into risk categories. Due to the rapid advance of omics technologies such as genomic, transcriptomics and proteomics, their use in precision medicine research has become common where the aim is to construct prediction models across a diverse range of data types. This increased complexity in turn can lead to poor model performance and biased results, which highlights the need for appropriate model validation. A major challenge that hinders the progress of precision medicine is lacking or incomplete validation for existing models, and difficult to define outcomes in the absence of a gold standard definition.

This project will use existing data from the MAximising Therapeutic Utility for Rheumatoid Arthritis (MATURA) project to predict therapeutic treatment response. MATURA is an observational study collecting genetic and clinical data on 5,300 patients with rheumatoid arthritis treated with either methotrexate, anti-tumour necrosis factor (anti-TNF) or rituximab. The aim of this project will be to apply machine learning and traditional statistical methods to develop and validate effective predictive modelling strategies through the integration of data available through MATURA and adjacent datasets. Available data includes phenotype data (demographics and clinical records), biomarker data (genetics, gene expression, metabolomics and epigenetics) and cellular immunophenotyping.