Despite the ever-increasing volume of high-resolution biological data, its analysis remains strongly grounded in first principles, as our goal is to gain insight into the underlying mechanisms that generate the data. Thus, the investigation of biological data often employs mechanistic mathematical models. Two main challenges in developing a mechanistic model are parametrisation (the forward problem) and calibration against the observed data (the inverse problem). Though these two tasks must be addressed on a case-by-case basis, there are common steps that can be taken to streamline the modelling process across various scenarios. In this talk, we will consider three case studies that highlight different aspects of model development for biological systems. The first study focuses on cellular biology, where a Hidden Markov Model is developed to infer hidden variables from the observed data. The second study, set in the microbiology domain, is an example of developing a dynamical model completely from first principles in anticipation of the experimental data against which the model will be calibrated. The third study, related to cardiac electrophysiology, involves building a dynamical model that incorporates experimentally observed ratios as its parameters. Through these examples, we will attempt to summarise common practices of model synthesis for biological systems. Additionally, we will discuss the challenges and advantages of engaging with the domain specialists during model development, methods of assessing the model quality, and examples of dealing with messy data.
|
 |
Over the last fifty years, the interdisciplinary area of nonlinear chemical dynamics has significantly expanded, characterised by a cooperative interplay between theory and experiment. These systems display a wide range of behaviours, including sudden shifts in stability and the emergence of intricate patterns that cannot be captured by linear mathematics alone. In chemical systems, such non-linearities often stem from mass action kinetics, particularly in reactions far from equilibrium. This research develops numerical kinetic models of enzyme systems to explore how system design, confinement, feedback, and mass transport influence biochemical rhythms and other interesting phenomena. \vspace{0.1em}
Bio: Anna is a fourth-year PhD student in Chemical, Biological and Materials Engineering at the University of Sheffield, supervised by Professor Annette Taylor. Her research focuses on numerical modelling of enzyme-based systems, exploring interesting dynamics through reaction kinetics and transport processes. She is part of a Leverhulme Trust-funded collaboration with the University of Leeds, with interests in integrating experiments and kinetic modelling for applications in artificial cells, materials science, drug delivery, and sensing.
|
|
|
When designing a clinical trial, the sample size must be carefully planned to ensure reliable results. If too few patients are enrolled, the trial may lack sufficient statistical power to demonstrate the treatment’s efficacy. Conversely, enrolling too many patients may expose them to unnecessary risks and increase costs. Traditionally, sample size determination relies on the statistical concept of ‘power,’ which assumes that the treatment has the expected effect. However, this assumption may not always hold. We introduce an alternative approach, the ‘Probability of Success’ method, which accounts for uncertainty in clinical trial design.
Bio: James Salsbury is a fourth year statistics PhD student at the University of Sheffield, supervised by Professors Jeremy Oakley, Steven Julious and Dr Lisa Hampson. His PhD is in collaboration with Novartis [a pharmaceutical company], looking at probability of success calculations for survival trials. His interests lie in Bayesian statistics, in particular eliciting prior distributions from experts and quantitative decision-making.
|
|
|
Metallic alloys are the cornerstone of modern industry, driving everything from transport in the aerospace and automotive industries, to energy generation, healthcare and beyond. The continuous development of new alloys is essential for improving performance in service and enabling future technologies. However, alloy development remains a long, costly, and iterative process with limited adaptability. Machine learning presents an opportunity to accelerate and reduce the cost of alloy discovery. In the first part of this talk, we’ll explore how machine learning can drive the discovery of novel hard metal alloys, providing a step change in the design cycle. Despite this, a key limitation in applying machine learning to alloy design is the lack of sufficiently large and high-quality datasets. While extensive experimental data exists in the scientific literature, it is often scattered and unstructured, making it difficult to leverage for computational analysis. To address this challenge, the second part of this talk will discuss leveraging large language models (LLMs) and natural language processing (NLP) to extract, curate, and construct alloy property databases directly from published research. Performing a comparative analysis between manually curated datasets and those generated through LLM-driven literature mining, evaluating the challenges and opportunities of automated data extraction for materials informatics. By integrating machine learning-driven alloy design with automated literature-based data extraction, this research aims to facilitate the development of comprehensive databases, ultimately accelerating alloy discovery and enhancing the effectiveness of computational materials design.
Bio: I am an experimental physicist by background, holding a master's degree in physics from the University of Sheffield before transitioning into material science. My PhD at the University of Sheffield focussed on materials informatics and alloy design, where I collaborated with the AIRE team in the computer science department, working at the intersection of material science and data-driven methodologies. Currently I am a postdoctoral research associate at the University of Sheffield, specialising in powder feedstock recycling and reconditioning to enhance the sustainability of industry manufacturing by reducing its environmental footprint.
|
|
|
What did you do at the end of the school day? Would you go to a sports club with friends? Or go to the cinema with your family? Or perhaps you needed to go home to cook and clean for a loved one? In this paper I look at exactly that, constructing a model of cultural participation for children with and without caring roles. Childhood cultural participation is linked to beneficial outcomes in one's educational attainment, employment and wider opportunities. Despite this, current literature is scarce of inquiry into the impacts of cultural participation on children. Additionally, young carers occupy a precarious position within the youth demographic, as their informal caring responsibilities take time and energy from other aspects of their lives. In this seminar, I will give a deep dive into my implementation of confirmatory factor analysis to construct a model of cultural participation of children. Further, I give critiques of PCA and aggregate measures to best justify the use of CFA. However, despite the value of exploring cultural participation for children with and without caring responsibilities using this methodology, we also need to be aware of its restrictions. With this in mind, I’ll also talk about data limitations I’ve experienced in building this model, and the drawbacks this has on building factors that entirely encompass different facets of cultural participation
Bio: I’m a PhD student in the Sheffield Methods Institute (SMI), researching outcomes of Young Carers. Through this, I’m very grateful to be partnered with Sheffield Young Carers, who have been amazing throughout! Previously, I’ve completed a Masters in Clinical Research and a Bachelors in Maths and Economics, both also from the University of Sheffield. Previous work as a mentor for young people, in addition to being part of student focused research projects has helped further in situating me in the Young Carer research space.
|
|
|
|
|
|