CfS Annual Conference 2025

The identification of biomarkers with high predictive accuracy is a crucial task in medical research, as it may help clinicians make early decisions, thus reducing morbidity and mortality in high-risk populations. Time-dependent receiver operating characteristic (ROC) curves are the main tool used to assess the accuracy of prognostic biomarkers for outcomes that evolve over time. Recognising the need to account for patient heterogeneity when evaluating the accuracy of a prognostic biomarker, we introduce a novel penalised-based estimator of the time-dependent ROC curve that accommodates the possible modifying effect of covariates. We consider flexible models for both the hazard function of the event time given the covariates and biomarker and for the location-scale regression model of the biomarker given covariates, enabling the accommodation of non-proportional hazards and nonlinear effects through penalised splines, thus overcoming limitations of earlier methods. Our approach is applied to evaluating the Global Registry of Acute Coronary Events risk score’s ability to predict mortality after discharge in patients who have suffered an acute coronary syndrome and how this ability may vary with the left ventricular ejection fraction. The results of a simulation study demonstrate that our approach successfully recovers the true functional form of the covariate-specific time-dependent ROC curve and the corresponding area under the curve across a variety of scenarios, and performs favourably compared to existing methods.

Synthetic biology aims to engineer biological systems with specified functions. This requires navigating an extensive design space, which is challenging to achieve with wet lab experiments alone. To expedite the design process, mathematical modelling is typically employed to predict circuit function in silico ahead of implementation, which when coupled with computational optimization can be used to automatically identify promising designs. However, circuit models are inherently inaccurate which can result in sub-optimal or non-functional in vivo performance. To mitigate this issue, here we propose to combine Bayesian inference, Thompson sampling, and risk management to find optimal circuit designs. Our approach employs data from non-functional designs to estimate the distribution of the model parameters and then employs risk-averse optimization to select design parameters that are expected to perform well given parameter uncertainty and biomolecular noise. We illustrate the approach by designing robust adaptation circuits and genetic oscillators with a prescribed frequency. The proposed approach provides a novel methodology for the design of robust genetic circuitry.

Dr Encarni Medina-Lopez is a Senior Lecturer in Ocean Observation at the School of Engineering, University of Edinburgh. In her presentation, Encarni will explore the significance of observing the oceans from space and address the current challenges associated with monitoring coastal waters. She will examine the range of available datasets, including both in situ measurements and satellite-derived data, and will discuss her research employing various machine learning techniques to generate high-accuracy ocean data products in complex and dynamic environments.  This will lead to the second part of her talk, which will focus on the intersection of data availability and climate justice, concluding with an introduction to her latest initiative: the Failure Modes of Engineering project.

Generative diffusion models are proving to be powerful tools for solving Bayesian inverse problems, delivering highly accurate posterior samples. However, their computational cost is often prohibitive due to the large number of neural evaluations needed per sample—especially in high-dimensional applications like computational imaging, where uncertainty quantification requires many samples. In this talk, I’ll present a Multilevel Monte Carlo framework tailored to diffusion-based samplers, which exploits the cost-accuracy trade-offs of these models to drastically cut computational costs. By coupling models of varying precision, we achieve the same final accuracy with up to 4x-8x lower computational effort, demonstrated across three benchmark imaging problems.

This presentation seeks to spark discussion over how we engage those who would rather avoid maths and or statistics in data analysis. The social sciences are a good case study for these issues as they encompass a range of mathematical capability, from high numeracy (e.g. economics) and low numeracy disciplines (e.g. anthropology). The presentation draws upon the research literature and findings from a National Centre for Research Methods project. In this, I will argue that there is weak evidence that fear of statistics has a meaningful impact on course performance, and that the evidence base itself is limited. The talk concludes with some suggestions and strategies of how to engage those who may be worried or hostile with statistical analysis.