We use cookies to ensure that we give you the best experience on our website. You can change your cookie settings at any time. Otherwise, we'll assume you're OK to continue.

Institute of Advanced Study

Past Events

Living Scales: Communicating Across Scales: Biophysics and Beyond - Multi-scale self-organised adaptive biological networks Workshop

8th February 2017, 09:00 to 17:00, Seminar Room, Institute of Advanced Study

Life is inherently multi-scaled ranging from molecules (nm) to 100m (trees) and from femtoseconds (chemical reactions) to millennia (evolution), covering around 27 orders of magnitude in total. Thus ‘Scale’ creates immediate interdisciplinary resonances within current novel attempts to make progress understanding complex biological systems, especially by bringing the natural and life sciences communities together. Scale in biophysical systems also brings serious interdisciplinary challenges:

i. How can we measure structure and dynamics simultaneously at different scales?

ii. How can we build models that capture multiscale phenomena?

iii. How can we create holistic, ‘systems’ understanding even if we possess (i) and (ii)?

Durham strengths lie in macromolecular biophysics (McLeish, Quinlan) and bio imaging (Girkin) with a strong emphasis on vision science. Inviting imaginative and creative international fellows (Fricker and Smithson) into the Institute of Advanced Study and the Biophysical Science Institute communities looking at scale more widely will be extremely productive.

The current, highly reductionist, methodologies deployed in the natural sciences are in transition as they attempt to increase parameter space and length scales by building from the smallest scales up. In the science of inanimate soft matter and statistical mechanics, we are accustomed to emergent behaviour from smaller scales to larger (e.g. viscous fluid behaviour from mutually attractive molecules). However, the behaviour of individual proteins, genes, cells, even single gene networks, turns out increasingly to be coupled not only to other entities at their own level, but essentially upwards and downwards in scale. So, for example, gene transcription and post-translational modification within a cell nucleus can depend on tissue-scale stress fields which in turn can be linked with whole body stress. Such effects are common across the whole spectrum of life being seen in both plants and animal. Scale-coupling like this is crucially involved in pathologies such as cardiomyopathy and cancer as well as more sociological induced stresses. The complexities opened up by the interference of bottom-up and top-down coupling have yet to be explored. A muchmore holistic approach to, and view of, scale is needed in biological systems. This constitutes the core of the proposed sub-theme.

The close engagement of biological and physical sciences across length-scales is not only a better way of providing insight into biology questions. It also teaches new ideas to the logic of statistical physics, mathematics of complex systems and the phenomenon of emergence. The reciprocal coupling across scales discussed above has a different character even from those non-linear couplings that give rise to chaotic dynamics and related phenomena, and will enrich a number of communities in Durham working on complex systems (e.g. the Tipping Points team, members of the EPSRC network on Complex and Non-Equilibrium Systems).

Experimental investigations of biophysical systems at different spatial and temporal scales present additional challenges such as the 27 orders of magnitude mentioned above. This is turn generates questions of visualisation of such complex data sets. Logarithmic scales provides a solution which can work in two simple dimensions (i.e. a graph) - but other ways of representing information and structures across length scales invite investigation when one has living systems which inherently operate in four dimensions. This has significant links through to the analysis of highly complex data in fields such as banking, cosmology, particle physics, seismology and marketing information thus with links to iARC.

These questions are generating very early-stage research indicating that there is a rich seam to mine. An obvious extension, focussed on Durham interests, would be to explore the biophysical basis of visual perception from the molecular-level operation of photoreceptors, through local information-processing neuronal networks in the retina to system-level neuroscience. This system creates regulation of visual sensitivity over the 10^12-fold variation in environmental light levels under which human vision operates. This vast dynamic range is maintained through sensitivity adjustments at all levels in the visual system, from individual photoreceptors, to retinal circuits, and complex neural coding in cortex. A second strategic topic to explore within this field is imaging and modelling of self-organised, adaptive networks that operate from a sub-cellular level (cytoskeleton and endoplasmic reticulum), through embedded transport networks (leaf venation and blood vascular systems), to entire organisms that grow as networks (fungi and slime moulds).

These central ideas are the core of three workshops held during the Epiphany term under the ‘Living Scale’ sub-theme. Each will explore in a completely interdisciplinary way, one living example of a multi-scale, coupled system.

Workshop 2 - 8th February 2017 – Cosin’s Hall, IAS
Multi-scale self-organised adaptive biological networks

Physical transport networks exist across a range of scales in biology from the endoplasmic reticulum at a sub-cellular level, to vascular systems in plants and animals, or even entire organisms in the case of fungi and slime molds. In each case there is a need to characterise the network architecture and behaviour over time, under different experimental conditions, or in normal and pathological states. This workshop will explore first the image analysis approaches available to extract the 2-D and 3-D network architecture that are common across different domains; second what type of graph-theoretic measurements are useful in the biological domain to describe these networks; and third how to integrate quantitative experimental network measurements into mathematical models of transport systems.

Workshop Timetable 8 February 2017

Abstracts and Biographies

Super-resolution of sub-cellular networks

Tim Hawkins

Durham Centre for Bioimaging Technology, Department of Biosciences, Durham University

Vessel Enhancement in Biomedical Images Using Granulometric Approaches

Cigdem Sazak

Bioimage Informatics Lab, Durham University

Abstract: Vessel enhancement, particularly before the extraction of vascular structures, is a key stage in many automated diagnostic and clinical pipelines. Whilst a number of image processing techniques have been proposed for this, many still have issues, especially with regards to junctions and branches and the wide variety of scales involved in a biological network. In this paper we introduce a new approach to vessel enhancement based upon mathematical morphology. Mathematical morphology is widely used in image processing and has been used in image enhancement previously. However, the approach proposed here combines different structuring functions together to identify innate features of vessel-like objects. This new approach is robust at junctions and able to cope with variations in thickness and contrast throughout vessel-like structure. We evaluate the proposed approach with both synthetic and real data and compare the results quantitatively with several existing vessel enhancement methods. Our results show that this approach is promising with high-quality enhanced images, accuracy at junctions and robustness to noise.

Structure and Dynamics of ER: Minimal Networks and Biophysical Constraints

Congping Lin,

College of Engineering, Mathematics and Physical Sciences, University of Exeter

Abstract: The endoplasmic reticulum (ER) in live cells is a highly mobile network whose structure dynamically changes on a number of timescales. Previous studies have identified potential static elements that the ER may remodel around. Here, we use these structural elements to assess biophysical principles behind the network dynamics. By analyzing imaging data of tobacco leaf epidermal, we show that the geometric structure and dynamics of ER networks can be understood in terms of minimal networks. Our results show that the ER network is well modeled as a locally minimal-length network between the static elements that potentially anchor the ER to the cell cortex over longer timescales; this network is perturbed by a mixture of random and deterministic forces. The network need not have globally minimum length; we observe cases where the local topology may change dynamically between different Euclidean Steiner network topologies. Using a Langevin approach, we model the dynamics of the nonpersistent nodes and use this to show that the images can be used to estimate both local viscoelastic behavior of the cytoplasm and filament tension in the ER network. This means we can explain several aspects of the ER geometry in terms of biophysical principles.

Hierarchical Centerline Enhancement and Extraction of Curvilinear Structures in 2D and 3D Images

Shuaa S. Alharbi

Bioimage Informatics Lab, Durham University

Abstract: Curvilinear structures, or more simply sets of lines, act as important descriptors of image objects with many practical applications in medicine, engineering, computer vision, and the biosciences. The main problem with state-of-the-art extraction methods are that they introduce multiple parameters causing an issue of robustness. For example, well-tuned methods often fail to capture objects with differing object class, scale, or noise level. In this paper, we define a new objective function, which when minimized produces an ordered hierarchical sequence of curvilinear structures capturing the largest and most influential paths first, and then progressively extracting smaller and smaller lines, i.e. until the level of imaging noise is reached. The objective function is minimized efficiently with stochastic optimization, where we show our results using direct Monte Carlo sampling. We derive analytical solutions to the processes without requiring user-tunable parameters, and demonstrate our results both quantitively and qualitatively. The method is shown to outperform state-of-the-art methods in certain cases of noise, object class, and scale, while remaining fundamentally easier to use.

Using Convolutional Neural Networks to segment network images

Hao Xu, Institute for Biomedical Engineering, Oxford

Abstract: Neural networks (NNs) or NN-like models have been around for decades, and the key learning algorithm, backpropagation (BP), was developed since 1960s. Until the 1980s, although in principle that BP allowed for deep networks, additional hidden layers often did not seem to have empirical benefits. There was a turning point for large-scale object recognition in the 2012 ImageNet Large Scale Visual Recognition Challenge (ILSVRC2012), when large-scale deep Convolutional Neural Network (CNN) from the SuperVision team entered the challenge and made deep learning (DL) popular.

There were a wide range of successful applications of DL in recent years, and the most famous one should be the gaming artificial intelligence (AI), Alpha-Go, from Google DeepMind. It included supervised learning (SL), reinforcement learning (RL) and the idea of transfer learning. As for the extraction of network architecture, our application of CNN for leaf vein segmentation also appears to have a good performance, while the preparation of the training data is critical. Fully convolutional network (FCN) has a variety of benefits and the potential for future work.


Mr Hao Xu received his undergraduate degree in Engineering Science in 2015, and started a PhD in Medical Image Analysis in the same year. His research interest is in biomedical image analysis, with a particular focus on the application and development of machine learning algorithms. In addition to research in the field of biomedical imaging, he has also worked in segmentation of network architecture.

Graph Theoretic Approaches to Network Analysis


Bioinspired rules for adaptive networks

Mark Fricker,

Department of Plant Sciences, University of Oxford

Abstract: The plasmodium of the large, single-celled amoeboid slime mold Physarum polycephalum is able to construct hydrodynamically optimized vein-networks that are capable of solving simple geometric problems. We have developed tools to extract the network architecture from time-series of Physarum development using intensity-independent, phase-congruency edge enhancement and watershed segmentation to give a set of weighted adjacency matrices containing the physical dimensions of each vein. In plasmodia withdrawing from an arena via a single exit, the change in volume for each vein and intervening plasmodial sheet can be used to predict the net flow across the network. A linear relationship was found between predicted flow and vein radius, consistent with predictions from Murray's Law. Furthermore, we show that mathematical models for self-organised, adaptive transport in Physarum simulate the experimental network organisation well if the scaling coefficient of the current-reinforcement rule is set to 3. In simulations, this resulted in rapid development of an optimal network that minimised the combined volume and frictional energy in comparison with other scaling coefficients. The simplicity of the bioinspired Physarum model hints at a class of algorithms that give quasi-optimal solutions to balancing cost and transport efficiency using a combination of iterative local rules with long-range coupling.


Prof. Mark Fricker read Botany at Oxford and undertook his doctorate with Colin Willmer at Stirling on dissecting signal transduction pathways in stomatal physiology. This led to the development of measurements of guard cell Ca2+ dynamics as a post-doc with Tony Trewavas in Edinburgh, which evolved into the current interest in quantitative imaging of Ca2+, pH and redox signalling and nutrient transport in networked systems that he has pursued in Oxford since 1989. His research cover a range of scales including confocal ratio imaging on a micron scale, radiolabel scintillation imaging at an intermediate scale, and network analysis and mathematical modelling to predict behaviour at a macro-scale. He is currently an IAS Fellow at Durham working with Boguslaw Obara to develop high-throughput, automated network analysis techniques.

Growth and control in fungal networks

Luke Heaton,

Department of Plant Sciences, University of Oxford and Mathematics, Imperial College

Abstract: Filamentous fungi are not unicellular, but unlike other multi-cellular organisms, they grow as physiologically integrated networks with no fixed morphology or persistent functional specialization (apart from the reproductive structures). In particular, saprotrophic woodland fungi form extensive, macroscopic networks that connect patches of ephemeral resources. These organisms continuously adapt to local nutritional or environmental cues through a limited developmental repertoire of growth, branching, fusion and regression. Their fitness depends on their ability to find and colonise patches of resource while remaining connected in the face of damage and predation, but the coordination of nutrient transport and the developmental logic of fungal networks remain profoundly mysterious. Through a combination of advanced imaging, radio-labelling and mathematical modelling, we have shown that the pattern of long-distance transport can be explained by the biophysical concept of growth-induced mass flow, whereby currents flow towards the growing regions. This mechanism for global control is possible because fungal cytoplasm is relatively free to flow across fungal colonies, so hydrostatic pressure equalizes rapidly, and any local mechanisms for controlling turgor and water uptake will necessarily have a global effect. Furthermore, fungal growth and the movement of tracer are often observed to transition from small, symmetric colonies to asymmetric growth through spontaneous symmetry breaking or response to resource discovery. Models and experiments suggest that hydraulic coupling may be a key mechanism for colony wide coordination, enabling globally efficient transport and foraging behaviour in growing networks with no central control.


Dr Luke Heaton has been studying fungal networks for 8 years, obtaining a doctorate in “Biological Transport Networks” from the University of Oxford, Department of Physics in 2012. Since then he has worked as a post-doc in the Department of Plant Sciences, University of Oxford, and the Department of Bioengineering, Imperial College. As well as conducting original research, Dr Heaton is interested in writing popular science, and his book, “A Brief History of Mathematical Thought”, has been published by Robinson and Oxford University Press. His research is currently funded by a grant from the Leverhulme Trust.

The interplay between global and local cellular interactions regulates morphogenesis

Prof. George Bassel

School of Biosciences, University of Birmingham

Abstract: Life originated as single celled organisms, and multicellularity arose multiple times across evolutionary history. Following this major transition, a progressive drive towards increasing complexity in multicellular configuration has occurred. This has enabled organisms to adopt novel functions and fill previously uninhabited niches. Uncovering the properties of these synergistic cellular configurations is central to identifying these optimized organizational principles, and to establish the structure-function relationships underlying complex organ design. We have developed methods to capture all cellular associations within plant organs using a combination of high resolution 4D microscopy and computational image analysis. These multicellular organs are abstracted into cellular connectivity networks and the organizational properties of their assemblies can be revealed using network science. Using this approach, we examined the topological dynamics of cellular organization within the apical stem cell niche of growing plants. The ability to predict when and how a cell divides was provided through a global path length topological proxy. This provides evidence for a previously undescribed mobile global signal moving across this organ. The flux of this signal through the intercellular transport network in turn influences the fate of individual cells within the community. This supports a model whereby each local and global signals regulate multicellular homeostasis in complex plant organs.


Prof. George Bassel is a Chair in Plant Computational Biology at the University of Birmingham. His research seeks to understand complexity in biological systems, and how this shapes and constrains life. This is being studied primarily across two scales: 1) Molecular complexity – how interactions between molecular components give rise to cellular behaviours. 2) Cellular complexity – how interactions between cells influence multicellular organ function. He is using plants as a model system to address these questions. The cells within plant organs are fixed together such that there is no cell migration, as in animal systems. Information moves through these immobilized cellular networks through shared cell walls. These cellular networks can be captured in 3D at high resolution and analysed. This research lies at the interface between biological, computational and mathematical sciences.

Contrast-Independent Curvilinear Structure Enhancement in 3D Biomedical Images

Boguslaw Obara

School of Engineering and Computing Sciences, Durham University

Abstract: A wide range of biomedical applications require detection, quantification and modelling of curvilinear structures in 3D images. Here we propose a 3D contrast-independent approach to enhance curvilinear structures based on the 3D Phase Congruency Tensor concept. The results show that the proposed method is insensitive to intensity variations along the 3D curve, and provides successful enhancement within noisy regions. The quality of the 3D Phase Congruency Tensor is evaluated by comparing it with state-of-the-art intensity-based approaches on both synthetic and real biological images.


Boguslaw Obara is a Lecturer in the School of Engineering and Computing Sciences, Durham

University, UK. His research focuses on development of image informatics approaches in order to fully automate the processing, quantification, and analysis of multi-dimensional images obtained by a wide spectrum of micro and macro imaging modalities, including AFM, transmitted light microscopy, fluorescence (confocal, multiphoton, total internal reflection, FRET, lifetime imaging, super-resolution) and macro-photography. His research activities are in fields ranging from science, engineering, to arts and humanities.

For further information about this workshop contact Dr Boguslaw Obara (; Professor Mark Fricker (visiting IAS Fellow).

To book on the Workshop, 8th February 2017, please follow this link;

A further two workshops will take place in this series:

This workshop series is sponsored by the Institute of Advanced Study and Biophysical Sciences Institute.

Contact for more information about this event.