Personalised Therapeutics

In recent years, AI and machine learning have been revolutionising healthcare in remarkable ways. Transforming various aspects of the field, ranging from diagnosis and treatment to personalised care and drug discovery. By leveraging vast amounts of data and powerful computational algorithms, AI and ML enable healthcare professionals to make more accurate predictions, improve patient outcomes, and streamline processes.

The van der Schaar lab is finding itself at the forefront of these developments working tirelessly on pushing boundaries and accelerating progress with the release of cutting-edge open-source software packages on the regular.

Personalised medicine, an approach that tailors medical treatments to individual patients, has been significantly impacted by the advancements of AI and ML technologies already. It comes as no surprise that it takes a prime spot on our lab’s research agenda.

Personalised Treatment – Overview

Current treatment guidelines have been developed with the “average” patient in mind (on the basis of randomised control trials), but there is ample evidence that different treatments result in different effects and outcomes from one individual to another: for any given treatment, it is quite likely that only a small proportion of people will actually respond in a manner that resembles the “average” patient.

Rather than making treatment decisions based on blanket assumptions about “average” patients, the goal of clinical decision-makers is increasingly shifting to determining the optimal treatment course for any given patient at any given time.

Read more about Individualised treatment effect inference in our dedicated Research Pillar here.

Our lab has built a position of leadership in this area. We have defined the research agenda by outlining and addressing key complexities and challenges, and by laying the theoretical groundwork for model development. In our development of algorithms, we have identified and targeted an extensive range of potential clinical applications using both clinical trials and observational data as inputs.

We put together a page with 6 tutorials on individualised treatment effect inference, each of which takes a different approach to the topic. You can find them here.

The first step on the way to a more personalised medicine is to change the methodology used in the formative stages of drug development.

Next-Generation Clinical Trials

Clinical trials are complex projects involving multiple data sources, design choices, and analytical methods. Some of these aspects are detailed in our research pillars: individualised treatment effect, clustering, survival analysis, etc.

Next-generation clinical trial design and implementation is one of our lab’s key research priorities, and our lab has developed an array of novel approaches. Our most recent ICML 2023 paper, for example, investigates how to incorporate new developments in ML and how to address challenges when designing algorithmic solutions.

Read more about Next-Generation Clinical Trials and find our publications on the topic in our dedicated Research Pillar here.

In a recent episode of our Revolutionizing Healthcare engagement series, we approached leading clinical experts to thoroughly discuss the current state of clinical trials and how ML can improve them in many ways. You can find a comprehensive write up here. This session was followed up by exploring the topic from a different perspective during an Inspiration Exchange roundtable with industry experts. You can find that one here.

Machine Learning meets Pharmacology

Clinical pharmacologists understand that patients are unique biological systems that respond differently to drugs and that adjusting dosing to specific circumstances is often necessary. In everyday bedside work as in clinical trials, trial and error can be time-consuming, and waiting weeks to learn how a patient responds is not always feasible. That is when mathematical models come in to help make predictions about patient outcomes.

However, what model should be used for the situation at hand?

In a new ICML 2023 publication, we try to entangle the ever-growing ML toolbox to make it easier for clinicians and pharmacologists to actually find and use the most appropriate tool. You can find it here.

Adapting existing methods and constructing new model ensembles for novel cases can be done robustly with our novel Synthetic Model Combination (SMC) tool. Population pharmacokinetic models are often trained on certain demographic groups given the studies that are designed for data collection. For a new patient who does not necessarily fit into one of the existing demographics, different models may be more or less relevant and accurate. Naive ensembles ignore this fact and always incorporate evenly the predictions of each model, SMC on the other hand aims to up-weight the models that would appear to be more relevant.

To introduce this tool we have created a dedicated post including an insightful recorded conversation between Prof Richard Peck (CCAIM) and our PhD student Alex Chan. You can find those here.

Also just published has been a novel method dealing with informatively recorded observation in ongoing treatments. TESAR-CDE utilises Neural CDEs to learn treatment outcomes in the presence of informative rather than random sampling. Read the publication here.

You can find a comprehensive overview of our work intersecting with pharmacology here.

A full list of our papers on this and related topics can be found here.