van der Schaar Lab

Revolutionizing Healthcare: ML and Multi-omics

On 25 March, we discussed multi-omics and machine learning in a new series of Core Concept sessions for our Revolutionizing Healthcare program.

“Omics” is a common term in biology, referring to the comprehensive study of a particular class of biological molecules, usually on a large scale. For instance, proteomics denotes the study of the entire set of proteins expressed by an organism. Multi-omics goes further by integrating diverse biological data types, such as genomics, transcriptomics, proteomics, metabolomics, and more, to provide a comprehensive view of an organism’s biological functions. By weaving together this tapestry of data, ML can not just accelerate research; it’s enabling us to envision personalised medicine in ways we never thought possible.

The approach of Multi-omics has become very popular. It allows you to interrogate each of these aspects of the disease process, and by integrating them, it delivers insights into causal factors or mechanisms of disease.

Prof Andres Floto

During our session, which is part of our foundational Core Concept series, we explored the latest advancements in ML and multi-omics research and examined possibilities for synergies between the fields.

Prof Andres Floto (CCAIM & University of Cambridge) expertly introduced this topic in a straightforward manner. We are delighted to share his introductory talk with you:

Equipped with a solid understanding of multi-omics, let us shift our focus to machine learning tools. These tools not only aid in exploring complex data but also open up new use-cases. Dr Fergus Imrie, a postdoc at the van der Schaar Lab, elaborated on our lab’s work and shed light on recent advancements in machine learning for multi-omics.

You can watch him explain ML methods for Polygenic Risk Scoring, Biomarker discovery, Multi-omics integration, and more:

To take us even further, we hosted world-leading experts from Harvard and Heidelberg (Prof Kun Hsing-Yu and Prof Julio Saez-Rodriguez) who delivered captivating talks on their latest research and engaged in a thorough discussion with both the panel and the audience. Watch the full episode here:

If you would like to explore our work on ML for multi-omics, you can find our publications here. In addition, we have a dedicated page on Self-supervised, semi-supervised, and multi-view learning where we explain a) what sets multi-omics apart from other domains where self-supervised learning has been successful, and b) how these challenges can be overcome.

If you want to learn even more about ML and multi-omics, join us for the world’s first Machine Learning summer school tailored to clinicians, researchers, medical students, and industry professionals where we can go into even greater depth.

To participate in future sessions of Revolutionizing Healthcare, please sign up here if you are a practising clinician. You can also view past sessions in the archive.