Forthcoming events in this series
The Omicron BA.1 variant of SARS-CoV-2 was more transmissible and less severe than the preceding Delta variant, including in hosts without previous infection or vaccination. To investigate why this was the case, we conducted in vitro replication experiments in human nasal and lung cells, then constructed and fitted ODE models of varying levels of complexity to the data, using Markov chain Monte Carlo methods. Our results fitting a simple model suggest that the basic reproduction number and growth rate are higher for Omicron in nasal cells, and higher for Delta in lung cells. As growth in nasal cells is thought to correspond to transmissibility and growth in lung cells is thought to correspond to severity, these results are consistent with epidemiological and clinical observations. We then fitted a more complex model, including different virus entry pathways and the immune response, to the data, to understand the mechanisms leading to higher infectivity for Omicron in nasal cells. This work paves the way for using within-host mathematical models to analyse experimental data and understand the transmission potential of future variants.
While presenting the results of this study, I will use them to open a wider discussion on common problems in mathematical biology, such as the situations in which complex models are preferable to simpler models; when it is appropriate to fix model parameters; and how to present results which are contingent on unidentifiable parameters.
Image-based readouts of biology are information-rich and inexpensive. Yet historically, bespoke data collection methods and the intrinsically unstructured nature of image data have made these assays difficult to work with at scale. This presentation will discuss advances made at Recursion to industrialise the use of cellular imaging to decode biology and drive drug discovery. First, the use of deep learning allows the transformation of unstructured images into biologically meaningful representations, enabling a 'map of biology' relating genetic and chemical perturbations to scale drug discovery. Second, building such a map at whole-genome scale led to the discovery of a "proximity bias" globally confounding CRISPR-Cas9-based functional genomics screens. Finally, I will discuss how publicly-shared resources from Recursion, including the RxRx3 dataset and MolRec application, enable downstream research both on cellular images themselves and on deep learning-derived embeddings, making advanced image analysis more accessible to researchers worldwide.
Data-driven modeling and deep learning present powerful tools that are opening up new paradigms and opportunities in the understanding, discovery, and design of soft and biological materials. I will describe our recent applications of deep representational learning to expose the sequence-function relationship within homologous protein families and to use these principles for the data-driven design and experimental testing of synthetic proteins with elevated function. I will then describe an approach based on latent space simulators to learn ultra-fast surrogate models of protein folding and biomolecular assembly by stacking three specialized deep learning networks to (i) encode a molecular system into a slow latent space, (ii) propagate dynamics in this latent space, and (iii) generatively decode a synthetic molecular trajectory.
Abstract: Neurotransmission at chemical synapses relies on the calcium-induced fusion of synaptic vesicles with the presynaptic membrane. The distance of the vesicle to the calcium channels determines the fusion probability and consequently the postsynaptic signal. After a fusion event, both the release site and the vesicle undergo a recovery process before becoming available for reuse again. For all these process components, stochastic effects are widely recognized to play an important role. In this talk, I will present our recent efforts on how to describe and structurally understand neurotransmission dynamics using stochastic modeling approaches. Starting with a linear reaction scheme, a method to directly compute the exact first- and second-order moments of the filtered output signal is proposed. For a modification of the model including explicit recovery steps, the stochastic dynamics are compared to the mean-field approximation in terms of reaction rate equations. Finally, we reflect on spatial extensions of the model, as well as on their approximation by hybrid methods.
References:
- A. Ernst, C. Schütte, S. Sigrist, S. Winkelmann. Mathematical Biosciences, 343, 108760, 2022.
- A. Ernst, N. Unger, C. Schütte, A. Walter, S. Winkelmann. Under Review. https://arxiv.org/abs/2302.01635
The overall goal of the Sherwood lab is to advance genomic and precision medicine applications through high-throughput, multi-disciplinary science. In a shortened talk this past autumn, I described our recent efforts using combined analysis of rare coding variants from the UK Biobank and genome-scale CRISPR-Cas9 knockout and activation screening to improve the identification of genes, coding variants, and non-coding variants whose alteration impacts serum LDL cholesterol (LDL-C) levels.
In this talk, I will discuss our emerging efforts to optimize and employ precision CRISPR techniques such as base editing and prime editing to better understand the impacts of coding and non-coding variation on serum LDL-C levels and coronary artery disease risk. This work involves the development of novel high-throughput screening platforms and computational analysis approaches that have wide applicability in dissecting complex human disease genetics.
The term cancer covers a multitude of bodily diseases, broadly categorised by having cells which do not behave normally. Cancer cells can arise from any type of cell in the body; cancers can grow in or around any tissue or organ making the disease highly complex. My research is focused on understanding the specific mechanisms that occur in the tumour microenvironment via mathematical and computational modelling. In this talk I shall present a 3D individual-based force-based model for tumour growth and development in which we simulate the behaviour of, and spatio-temporal interactions between, cells, extracellular matrix fibres and blood vessels. Each agent is fully realised, for example, cells are described as viscoelastic sphere with radius and centre given within the off-lattice model. Interactions are primarily governed by mechanical forces between elements. However, as well as he mechanical interactions we also consider chemical interactions, by coupling the code to a finite element solver to model the diffusion of oxygen from blood vessels to cells, as well as intercellular aspects such as cell phenotypes.
Nanopore sequencing is a method to infer the sequence of nucleotides in DNA or RNA molecules from small variations in ionic current during transit through a nanoscale pore. We will give an introduction to nanopore sequencing and some of its applications and then explore simple models of the signal generation process. These can provide insight to guide optimisation of the system and inform the design of more flexible neural network models, capable of extracting the rich contextual information required for accurate sequence inference.
The Turing pattern is a key concept in the modern study of reaction-diffusion systems, with Turing patterns proposed a possible explanation for the spatial structure observed in myriad physical, chemical, and biological systems. Real-world systems are not always so clean as idealized Turing systems, and in this talk we will take up the case of more messy reaction-diffusion systems involving explicit space or time dependence in diffusion or reaction terms. Turing systems of this nature arise in several applications, such as when a Turing system is studied on a growing substrate, is subjected to a temperature gradient, or is immersed within a fluid flow. The analysis of these messy Turing systems is not as straightforward as in the idealized case, with the explicit space or time dependence greatly complicating or even preventing most standard routes of analysis. Motivated by patterning in phenomena involving explicit space or time dependence, and by the interesting mathematical challenges inherent in the study of such systems, in this talk we consider the following questions:
* Is it possible to obtain generalizations of the Turing instability conditions for non-autonomous, spatially heterogeneous reaction-diffusion systems?
* What can one say about predicting nascent patterns in these systems?
* What role does explicit space or time dependence play in selecting fully developed patterns in these systems?
* Is it possible to exploit this space or time dependence in order to manipulate the form of emergent patterns?
We will also highlight some of the applications opened up by the analysis of heterogeneous Turing systems.
The dynamics of a tissue in development or regeneration arises from the behaviour of its constituent cells and their interactions. We use mathematical models and inference from experimental data to to infer the likely cellular behaviours underlying changing tissue states. In this talk I will show examples of how we apply canonical birth-death process models to novel experimental data, how we are extending such models with volume exclusion and multistate dynamics, and how we attempt to more generally learn cell-cell interaction models directly from data in interpretable ways. The applications range from in vitro models of embryo development to in vivo blood regeneration that is disrupted with ageing.
Mathematical models of childhood diseases are often fitted using deterministic methods under the assumption of homogeneous contact rates within populations. Such models can provide good agreement with data in the absence of significant changes in population demography or transmission, such as in the case of pre-vaccine era measles. However, accurate modeling and forecasting after the start of mass vaccination has proved more challenging. This is true even in the case of measles which has a well understood natural history and a very effective vaccine. We demonstrate how the dynamics of homogeneous and age-structured models can be similar in the absence of vaccination, but diverge after vaccine roll-out. We also present some fundamental differences in deterministic and stochastic methods to fit models to data, and propose techniques to fit long term time series with imperfect covariate information. The methods we develop can be applied to many types of complex systems beyond those in disease ecology.
Approaches to personalized diagnosis and treatment in oncology are heavily reliant on computer models that use molecular and clinical features to
characterize an individual patient’s disease. Most of these models use genome and/or gene expression sequences to develop classifiers of a patient’s
tumor. However, in order to fully model the behavior and therapy response of a tumor, dynamic models are desirable that can act like a Digital Twin of
the cancer patient allowing prognostic and predictive simulations of disease progression, therapy responses and development of resistance. We are
constructing Digital Twins of cancer patients in order to perform dynamic and predictive simulations that improve patient stratification and
facilitate the design of individualized therapeutic strategies. Using a hybrid approach that combines artificial intelligence / machine learning
with dynamic mechanistic modelling we are developing a computational framework for generating Digital Twins. This framework can integrate
different types of data (multiomics, clinical, and existing knowledge) and produces personalized computational models of a patient’s tumor. The
computational models are validated and refined by experimental work and in retrospective patient studies. We present some of the results of the dynamic
Digital Twins simulations in neuroblastoma. They include (i) identification on non-MYCN amplified high risk patients; (ii) prediction of individual
patients’ responses to chemotherapy; and (iii) identification of new drug targets for personalized therapy. Digital Twin models allow the dynamic and
mechanistic simulation of disease progression and therapy response. They are useful for the stratification of patients and the design of personalized
therapies.
Reaction-diffusion waves have long been used to describe the growth and spread of populations undergoing a spatial range expansion. Such waves are generally classed as either pulled, where the dynamics are driven by the very tip of the front and stochastic fluctuations are high, or pushed, where cooperation in growth or dispersal results in a bulk-driven wave in which fluctuations are suppressed. These concepts have been well studied experimentally in populations where the cooperation leads to a density-dependent growth rate. By contrast, relatively little is known about experimental populations that exhibit a density-dependent dispersal rate.
Using bacteriophage T7 as a test organism, we present novel experimental measurements that demonstrate that the diffusion of phage T7, in a lawn of host E. coli, is hindered by steric interactions with host bacteria cells. The coupling between host density, phage dispersal and cell lysis caused by viral infection results in an effective density-dependent diffusion rate akin to cooperative behavior. Using a system of reaction-diffusion equations, we show that this effect can result in a transition from a pulled to pushed expansion. Moreover, we find that a second, independent density-dependent effect on phage dispersal spontaneously emerges as a result of the viral incubation period, during which phage is trapped inside the host unable to disperse. Our results indicate both that bacteriophage can be used as a controllable laboratory population to investigate the impact of density-dependent dispersal on evolution, and that the genetic diversity and adaptability of expanding viral populations could be much greater than is currently assumed.
Cell-to-cell variability is often a primary source of variability in experimental data. Yet, it is common for mathematical analysis of biological systems to neglect biological variability by assuming that model parameters remain fixed between measurements. In this two-part talk, I present new mathematical and statistical tools to identify cell-to-cell variability from experimental data, based on mathematical models with random parameters. First, I identify variability in the internalisation of material by cells using approximate Bayesian computation and noisy flow cytometry measurements from several million cells. Second, I develop a computationally efficient method for inference and identifiability analysis of random parameter models based on an approximate moment-matched solution constructed through a multivariate Taylor expansion. Overall, I show how analysis of random parameter models can provide more precise parameter estimates and more accurate predictions with minimal additional computational cost compared to traditional modelling approaches.
Building computationally capable biological systems has long been an aim of synthetic biology. The potential utility of bio-computing devices ranges from biosafety and environmental applications to diagnosis and personalised medicine. Here we present work on the design of bacterial computers which use spatial patterning to process information. A computer is composed of a number of bacterial colonies which, inspired by patterning in embryo development, communicate using diffusible morphogen-like signals. A computation is programmed into the overall physical arrangement of the system by arranging colonies such that the resulting diffusion field encodes the desired function, and the output is represented in the spatial pattern displayed by the colonies. We first mathematically demonstrate the simple digital logic capability of single bacterial colonies and show how additional structure is required to build complex functions. Secondly, inspired by electronic design automation, an algorithm for designing optimal spatial circuits computing two-level digital logic functions is presented, extending the capability of our system to complex digital functions without significantly increasing the biological complexity. We implement experimentally a proof-of-principle system using engineered Escherichia coli interpreting diffusion fields formed from droplets of an inducer molecule. Our approach will open up new ways to perform biological computation, with applications in synthetic biology, bioengineering and biosensing. Ultimately, these computational bacterial communities will help us explore information processing in natural biological systems.
The overall goal of the Sherwood lab is to advance genomic and precision medicine applications through high-throughput, multi-disciplinary science. In this talk, I will review a suite of high-throughput genomic and cellular perturbation platforms using CRISPR-based genome editing that the lab has developed to improve our understanding of genetic disease, gene regulation, and genome editing outcomes.
This talk will focus on recent efforts using combined analysis of rare coding variants from the UK Biobank and genome-scale CRISPR-Cas9 knockout and activation screening to improve the identification of genes, coding variants, and non-coding variants whose alteration impacts serum LDL cholesterol (LDL-C) levels. Through these efforts, we show that dysfunction of the RAB10 vesicle transport pathway leads to hypercholesterolemia in humans and mice by impairing surface LDL receptor levels. Further, we demonstrate that loss of function of OTX2 leads to robust reduction in serum LDL-C levels in mice and humans by increasing cellular LDL-C uptake. Finally, we unveil an activity-normalized base editing screening framework to better understand the impacts of coding and non-coding variation on serum LDL-C levels, altogether providing a roadmap for further efforts to dissect complex human disease genetics.
Cell fate decision-making is responsible for development and homeostasis, and is dysregulated in disease. Despite great promise, we are yet to harness the high-resolution cell state information that is offered by single-cell genomics data to understand cell fate decision-making as it is controlled by gene regulatory networks. We describe how we leveraged joint dynamics + genomics measurements in single cells to develop a new framework for single-cell-informed Bayesian parameter inference of Ca2+ pathway dynamics in single cells. This work reveals a mapping from transcriptional state to dynamic cell fate. But no cell is an island: cell-internal gene regulatory dynamics act in concert with external signals to control cell fate. We developed a multiscale model to study the effects of cell-cell communication on gene regulatory network dynamics controlling cell fates in hematopoiesis. Specifically, we couple cell-internal ODE models with a cell signaling model defined by a Poisson process. We discovered a profound role for cell-cell communication in controlling the fates of single cells, and show how our results resolve a controversy in the literature regarding hematopoietic stem cell differentiation. Overall, we argue for the need to consider single-cell-resolved models to understand and predict the fates of cells.
Recent advances in Deep Learning have enabled us to explore new application domains in molecular biology and drug discovery - including those driven by complex processes that defy analytical modelling. However, despite the combined forces of increased data, improving compute resources and continuous algorithmic innovation all bringing previously intractable problems into the realm of possibility, many advances are yet to make a tangible impact for life science discovery. In this talk, Dr Marcin J. Skwark will discuss the challenge of bringing machine learning innovation to tangible real-world impact. Following a general introduction of the topic, as well as newly available methods and data, he will focus on the modelling of COVID-19 variants and, in particular, the DeepChain Early Warning System (EWS) developed by InstaDeep in collaboration with BioNTech. With thousands of new, possibly dangerous, SARS-CoV-2 variants emerging each month worldwide, it is beyond humanities combined capacity to experimentally determine the immune evasion and transmissibility characteristics of every one. EWS builds on an experimentally tested AI-first computational biology platform to evaluate new variants in minutes, and is capable of risk monitoring variant lineages in near real-time. This is done by combining an AI-driven protein structure prediction framework with large, spike protein sequence-oriented Transformer models to allow for rapid simulation-free assessment of the immune escape risk and expected fitness of new variants, conditioned on the current state of the world. The system has been extensively validated in cooperation with BioNTech, both in terms of host cell infection propensity (including experimental assays of receptor binding affinity), and immune escape (pVNT assays with monoclonal antibodies and real-life donor sera). In these assessments, purely unsupervised, data-first methods of EWS have shown remarkable accuracy. EWS flags and ranks all but one of the SARS-CoV-2 Variants of Concern (Alpha, Beta, Gamma, Delta… Omicron), discriminates between subvariants (e.g. BA.1/BA.2/BA.4 etc. distinction) and for most of the adverse events allows for proactive response on the day of the observation. This allows for appropriate response on average six weeks before it is possible for domain experts using domain knowledge and epidemiological data. The performance of the system, according to internal benchmarks, improves with time, allowing for example for supporting the decisions on the emerging Omicron subvariants on the first days of their occurrence. EWS impact has been notable in general media [2, 3] for the system's applicability to a novel problem, ability to derive generalizable conclusions from unevenly distributed, sparse and noisy data, to deliver insights which otherwise necessitate long and costly experimental assays.
Experimental biologists study diseases mostly through their abnormal molecular or cellular features. For example, they investigate genetic abnormalities in cancer, hormonal imbalances in diabetes, or an aberrant immune system in vascular diseases. Moreover, many diseases also have a mechanical component which is critical to their deadliness. Most notably, cancer kills typically through metastasis, where the cancer cells acquire the capability to remodel their adhesions and to migrate. Solid tumours are also characterised by physical changes in the extracellular matrix – the material surrounding the cells. While such physical changes are long known, only relatively recent research revealed that cells can sense altered physical properties and transduce them into chemical information. An example is the YAP/TAZ signalling pathway that can activate in response to altered matrix mechanics and that can drive tumour phenotypes such as the rate of cell proliferation.
Systems-biology models aim to study diseases holistically. In this talk, I will argue that physical signatures are a critical part of many diseases and therefore, need to be incorporated into systems-biology. Crucially, physical disease signatures bi-directionally interact with molecular and cellular signatures, presenting a major challenge to developing such models. I will present several examples of recent and ongoing work aimed at uncovering the relations between mechanical and molecular/cellular signatures in health and disease. I will discuss how blood vessel cells interact mechano-chemically with each other to regulate the passage of cells and nutrients between blood and tissue and how cancer cells grow and die in response to mechanical and geometrical stimuli.