M.Sc. Florian Borse defends his PhD thesis "Modelling Microbial Population Growth Across Spatial, Strain and Environmental Diversity" on Friday the 24th of October 2025 at 12 in the University of Helsinki Main building, Room F3017 (Fabianinkatu 33, 3rd floor). His opponent is Professor Leo Lahti (University of Turku) and custos Professor Ville Mustonen (University of Helsinki). The defence will be held in English.
The thesis of Florian Borse is a part of research done in the Department of Computer Science and in the Bioinformatics and Evolution group at the University of Helsinki. His supervisors have been Professor Ville Mustonen (University of Helsinki) and Collegium Researcher, Docent Johannes Cairns (University of Turku).
Modelling Microbial Population Growth Across Spatial, Strain and Environmental Diversity
Assessing the size of a population and its change is a ubiquitous task in biology. Typically, this is done by counting individual members, although often population size estimates can be retrieved from size-related characteristics of a population — for example, its spread across an environment. Deriving models to allow for such estimates carries the additional benefit of increasing understanding on how populations grow.
This understanding for population growth is itself required in a wide variety of fields. In the field of medicine, growth can be desired, as in the regrowth of tissues and organs, or undesired, as in the case of pathogens infecting a host organism. In the field of evolution, the growth dynamics of subpopulations of individuals carrying a certain allele are crucial in order to understand the evolutionary dynamics of the whole population. Population growth occurs thus in a wide array of contexts and scales across biology.
Here we focus on three population growth contexts: isogenic populations distributed across a shared environment, the relation between single-strain populations and multi-strain communities, and finally temporally changing environments across a long enough time span leading to evolutionary change of the populations. To understand growth across these systems we employ a hybrid approach using first machine learning (ML) models to gain insight of the processes at work. We then use these insights to both complement domain-specific approaches and inform explicit mechanistic models developed to characterise population growth.
In the case of populations distributed spatially across a shared environment, we first find that the impact of location in several growth stages is due to intrinsic population specific aspects and the impact of populations on their immediate environment, as they consume its local resources. We then show that diffusion of available resources plays a major role in later growth stages, as populations located closer to the available excess resource grow much longer than those in less favourable locations. Our ability to mechanistically describe such growth dynamics opens a future possibility for massively parallel growth experiments, which will enable studying eco-evolutionary dynamics of microbes within a shared, spatially coupled, environment.
For multi-strain communities, we find that a regression model can be devised to describe the growth characteristics of multi-strain communities from early growth characteristics of single-strain populations. Additionally, similar regression models uncover here a relation of rapidly diminishing returns for higher order interactions. This implies that single and pairwise experiments can capture most aspects of growth in the multi-strain community in the system studied.For population growth across changing environments, we find that phenotyping a population to certain antimicrobials can reveal its past, namely the time since exposure to that antimicrobial. We demonstrate this by combining ML and information theory to examine the mutual information carried by end-point phenotypic data on the whole antimicrobial exposure history, which is far higher than that obtained by chance.
Our findings within each growth context uncover essential aspects of population growth, in the form of insights derived from the ML approach and utilising these results to develop explicit mathematical models. The developed data-driven modelling approach is broadly applicable to better understand growth and its modifiers from large scale experimental data beyond the study systems of this thesis.
Availability of the dissertation
An electronic version of the doctoral dissertation will be available in the University of Helsinki open repository Helda at
Printed copies will be available on request from Florian Borse: