Machine Learning for Biomedical Data

The field of machine learning provides methodologies that are ideally suited to the task of extracting knowledge from data. Both statistical modelling and ML seek to build a mathematical description, a model, of the data and the underlying mechanism it represents; thus inevitably there is substantial overlap between the two. However, historically they differ in their rationale as follows. Statistical models start with an assumption about the underlying data distribution (e.g. Gaussian, Poisson). The focus is on inference; estimating the parameters of the statistical model that most likely gave rise to the observed data, and providing uncertainty bounds for these estimates. For ML, the focus is typically on prediction; without necessarily assuming a functional distribution for the data, a model that achieves optimal predictive performance is identified. It is this hypothesis-free approach that makes ML an attractive choice for dealing with complex data sets. While in traditional statistical modelling a hypothesis (model) is put forward and is then accepted/rejected depending on how consistent it is with the measured observations, ML methods learn this hypothesis directly from the training data set. In this course, we will review some of the useful way Machine learning is used in clinical research (for clinical end-use) and in industry Research and Development, ranging from early discovery to challenges emerging in clinical trials.

Many problems of interest to clinicians and pharmaceutical companies can be addressed using machine learning approaches. These problems include:

• molecular mechanisms of disease
• Co-morbidity and other factors
• design of signatures for the identification of potential responders to therapies
• analysis of molecular mechanisms of disease progression;

Industry (pharma/biotech)
• Classification of patients
• detection of toxicity at the in vitro stage
• analysis of molecular mechanisms of a drug action, including identification of primary targets and “reaction waves” caused by the initial impact; “target discovery”
• identification of additional diseases for which a drug can be potentially efficient. “drug repurposing” or “repositioning”

Sophisticated commercial software solutions have been developed to address these challenges, however, knowledge of basic methods is important both for optimization of these solutions, and for their efficient usage with a correct and deep interpretation of results.