The Mathematical & Computational Biology Forum

A weekly 1-hour forum designed to bring together mathematical, computational and experimental biologists presenting current research. The key focus is to link theoretical models and data and provide the opportunity for new cross-disciplinary interactions. A novel and successful presentation format is employed, which maximises interaction with the diverse audience. Speakers are encouraged to discuss work in progress and we only allow a maximum of 3 slides to encourage interaction with the public and whiteboard usage. This has proven successful in increasing the involvement of the attendees. If you are looking for interaction with researchers on the areas of biology, mathematics or computer science, this is a perfect opportunity to receive diverse and extensive feedback and to find interdisciplinary collaborations. Please contact one of the organisers if you wish to contribute a talk. Please see below for dates and venues.

Organisers

Dr Rafał Bogacz - Department of Computer Science, r.bogacz (at) bristol.ac.uk.
Jakub Nowacki - Department of Engineering Mathematics, j.nowacki (at) bristol.ac.uk.

Former organisers:

Frank Marten - Department of Engineering Mathematics.
Alejo Nevado - Department of Computer Science.

Dates and venue

Our forum is held every Thursday during term time from 13:00 to 14:00 in MVB 4.01 (Merchant Venturers Building). On rare occasions we have to change the room, this will be announced in the table below. For the most recent updates, abstracts of talks and locations, please view our online calendar. We also have a mailing list to announce forum updates. If you wish to be added to this list, please contact Dr Rafal Bogacz.

Go to: February-May 2011 || November-December 2010

February-April 2011

10 February 2011	Marcus Kaiser
Slides:	PDF
Venue:	MVB 1.06
Title:	The human connectome: topological, spatial, and dynamic features of brain networks
Abstract:	The human brain consists of connections between neurons at the local level and of connections between brain regions at the global level. The study of the entire network, the connectome, has become a recent focus in neuroscience research. Using routines from physics and the social sciences, neuronal networks were found to show properties of scale-free networks, making them robust towards random damage, and of small-world systems leading to better information integration. I will describe the main features of the topological and spatial organisation of neural systems and how they differ from artificial systems information processing systems such as computers. Recent clinical studies in the last three years have shown that the network features of the healthy brain differ from that of schizophrenia, epilepsy, and Alzheimer’s disease patients. These features even differ depending on cognitive features such as IQ. I will show how network features and simulations of brain activity can be used to assess and model changes in patients. For example, simulating the spreading dynamics of epileptic seizures can inform of underlying reasons for epilepsy. I will finally outline how these feature extractions and models can be applied to clinical problems in neural networks.

17 February 2011	Jakub Nowacki
Title:	Intrinsic excitability of hippocampal pyramidal neurons: experiments and theory
Abstract:	Intrinsic excitability, in addition to synaptic transmission and network behaviour, could have a great impact on brain function. We have developed a unified CA1/3 pyramidal neuron model based on recent experimental results. A hallmark of excitability in hippocampal CA1/3 pyramidal neurons is the after-depolarising potential (ADP). We perform numerical analysis of ADP and the excitability in CA1/3 neurones that includes model parameter sensitivity analysis and continuation. In this talk I would like to focus on how modelling can help understanding experimental results and provide predictions for future experiments. Furthermore, I would like to discuss how experiments and theory converge in experimental techniques such as dynamic clamp.

17 February 2011

Jakub Nowacki

Title:

Intrinsic excitability of hippocampal pyramidal neurons: experiments and theory

Abstract:

Intrinsic excitability, in addition to synaptic transmission and network behaviour, could have a great impact on brain function. We have developed a unified CA1/3 pyramidal neuron model based on recent experimental results. A hallmark of excitability in hippocampal CA1/3 pyramidal neurons is the after-depolarising potential (ADP). We perform numerical analysis of ADP and the excitability in CA1/3 neurones that includes model parameter sensitivity analysis and continuation. In this talk I would like to focus on how modelling can help understanding experimental results and provide predictions for future experiments. Furthermore, I would like to discuss how experiments and theory converge in experimental techniques such as dynamic clamp.

24 February 2011	Mark van Rossum
Venue:	MVB 1.06
Title:	Weight dependent synaptic learning rules
Abstract:	The strength of the synapses in the brain are presumably continuously subject to increases and decreases as the result of ongoing learning processes. This realization allows one to approximate the synaptic weight evolution as a stochastic process. This has been used to find fundamental limits of storage. Recently we introduced a synaptic information capacity measure based on Shannon information (Barrett and van Rossum '08). We use this to find the optimal weight dependent learning rules. We find that soft-bound learning rules are better than hard bound rules, although the improvement is quite small. Furthermore, we show how feedforward inhibition further increases storage.

3 March 2011	John McGonigle
Slides:	PDF
Title:	Methodological issues and interpretation of functional connectivity in pharmacological MRI
Abstract:	We have recently been looking at data-driven approaches examining functional connectivity in human pharmacological MRI. Here we will discuss some of these methods and their issues, especially how changes in laterality should be measured and the form some functional networks take. Also, how these results should be interpreted at the relatively low resolution available with this type of functional imaging.

13 April 2011	Anil Seth
Venue:	MVB 1.06
Title:	Measuring Consciousness: Causal density and integrated information
Abstract:	An outstanding challenge in neuroscience is to develop theoretically grounded and practically applicable quantitative measures of neural signals that are sensitive to conscious level. Such measures should be high for vivid alert conscious wakefulness, and low for unconscious states such as dreamless sleep, coma, and general anesthesia. I will describe recent progress in the development of measures of dynamical complexity, in particular ‘causal density’ and ‘integrated information’. These and similar measures capture in different ways the extent to which a system's dynamics are simultaneously differentiated and integrated. Because conscious scenes are distinguished by the same dynamical features, these measures are therefore good candidates for reflecting conscious level. After reviewing the theoretical background, I will present some simulation results demonstrating similarities and differences between the measures, as well as recent results and challenges in the practical application of the measures to empirically obtained data.

5 May 2011	Maciej Klemm
Slides:	PDF
Title:	Development of the Dynamic Microwave Imaging system for contrast-enhanced cancer imaging and brain neuroimaging
Abstract:	In this short presentation we will present a concept of a novel Dynamic Microwave Imaging (DMI) system. This system is being developed within a recently funded EPSRC Career Acceleration Fellowship project. Targeted clinical applications include nanoparticle contrast-enhanced cancer imaging and brain neuroimaging.

19 May 2011	Katy Turner
Title:	Evaluating complex models of sexually transmitted diseases - some different approaches
Abstract:	Computational power is no longer a limiting factor in determining the complexity of models which can be used to simulate the dynamics of infectious disease. In recent years there has been a proliferation of models of varying complexity, but little help to decide which, if any, are "good" or "better than the others" or even "good enough". In this talk I'll present some recent published and current ongoing work which has attempted to address this problem of model evaluation and comparison, when the data are imperfect and the systems modelled are highly complex, as is the case for the spread of sexually transmitted diseases, such as chlamydia on a sexual network. In the first example I'll show what happens when you directly compare models; by a combination of reprogramming, harmonising and brute force to understand in a specific case why different models give different answers. I'll also mention a more statistically rigorous method we are in the process of developing along with Dan Lawson, Caroline Colijn and Katy Robinson and Richard Everitt. The second approach using ABC (approximate bayesian computing) promises to alleviate some of the problems of model comparison and objectivity as well as future-proofing models to some extent.

19 May 2011

Katy Turner

Title:

Evaluating complex models of sexually transmitted diseases - some different approaches

Abstract:

Computational power is no longer a limiting factor in determining the complexity of models which can be used to simulate the dynamics of infectious disease. In recent years there has been a proliferation of models of varying complexity, but little help to decide which, if any, are "good" or "better than the others" or even "good enough". In this talk I'll present some recent published and current ongoing work which has attempted to address this problem of model evaluation and comparison, when the data are imperfect and the systems modelled are highly complex, as is the case for the spread of sexually transmitted diseases, such as chlamydia on a sexual network. In the first example I'll show what happens when you directly compare models; by a combination of reprogramming, harmonising and brute force to understand in a specific case why different models give different answers. I'll also mention a more statistically rigorous method we are in the process of developing along with Dan Lawson, Caroline Colijn and Katy Robinson and Richard Everitt. The second approach using ABC (approximate bayesian computing) promises to alleviate some of the problems of model comparison and objectivity as well as future-proofing models to some extent.

26 May 2011	Michael Ashby
Title:	The network architecture of recurrent circuits in the brain
Abstract:	The brain activity that underlies sensory perception is generated by networks of neurons that are connected by and communicate via synapses. These connections are formed and modified early in life, at a time when the brain is growing rapidly, in a process that is guided by sensory experience. During this developmental period, the recurrent networks that are ubiquitous in the sensory cortex emerge through the specific addition of new synaptic connections. Modeling of these networks, using Graph Theory, shows that they assume a non-random architecture. In particular, these networks have a high degree of clustering and short path lengths, characteristic of small-world connectivity. It is postulated that small world network arrangements in the brain may facilitate information processing but it is not clear how this is manifested in a physiological context. Integrating predictions from dynamic network analysis with physiological measurements will be an important step in revealing how the synaptic architecture of recurrent brain circuits determines the patterns of activity that permit sensory perception.

26 May 2011

Michael Ashby

Title:

The network architecture of recurrent circuits in the brain

Abstract:

The brain activity that underlies sensory perception is generated by networks of neurons that are connected by and communicate via synapses. These connections are formed and modified early in life, at a time when the brain is growing rapidly, in a process that is guided by sensory experience. During this developmental period, the recurrent networks that are ubiquitous in the sensory cortex emerge through the specific addition of new synaptic connections. Modeling of these networks, using Graph Theory, shows that they assume a non-random architecture. In particular, these networks have a high degree of clustering and short path lengths, characteristic of small-world connectivity. It is postulated that small world network arrangements in the brain may facilitate information processing but it is not clear how this is manifested in a physiological context. Integrating predictions from dynamic network analysis with physiological measurements will be an important step in revealing how the synaptic architecture of recurrent brain circuits determines the patterns of activity that permit sensory perception.

November-December 2010

4 November 2010	David Kelly
Title:	Inferring hidden Markov models from noisy data
Abstract:	Cutting edge experimental data is frequently noisy, making quantitative analysis a challenge. Hidden Markov modelling is a popular approach to this challenge. Here we present an alternative method for inferring Hidden Markov type models from data without the necessity of assuming on number of states. We demonstrate their use with Fluorescent Resonant Energy Transfer spectra of DNA conformational dynamics.

4 November 2010

David Kelly

Title:

Inferring hidden Markov models from noisy data

Abstract:

Cutting edge experimental data is frequently noisy, making quantitative analysis a challenge. Hidden Markov modelling is a popular approach to this challenge. Here we present an alternative method for inferring Hidden Markov type models from data without the necessity of assuming on number of states. We demonstrate their use with Fluorescent Resonant Energy Transfer spectra of DNA conformational dynamics.

11 November 2010	Colin Campbell
Title:	Novel data integration methods and their biomedical applications
Abstract:	Substantial quantities of data are being generated within the biomedical sciences and the successful integration of different types of data remains an important challenge. Data can include expression array, exon array, SNP array, aCGH and microRNA array in addition to resequencing data from Next Generation Sequencing studies. Without technical detail, we outline some of our research on devising novel data integration methods for unsupervised, semi-supervised and supervised learning with multiple types of disparate data. We discuss some interesting current application studies involving these data integration methods.

11 November 2010

Colin Campbell

Title:

Novel data integration methods and their biomedical applications

Abstract:

Substantial quantities of data are being generated within the biomedical sciences and the successful integration of different types of data remains an important challenge. Data can include expression array, exon array, SNP array, aCGH and microRNA array in addition to resequencing data from Next Generation Sequencing studies. Without technical detail, we outline some of our research on devising novel data integration methods for unsupervised, semi-supervised and supervised learning with multiple types of disparate data. We discuss some interesting current application studies involving these data integration methods.

18 November 2010	Jack Mellor
Title:	A Ca2+-based computational model for NMDA receptor-dependent synaptic plasticity
Abstract:	Synaptic plasticity requires coincident activity in presynaptic and postsynaptic neurons to activate NMDA receptors (NMDARs). The resultant Ca2+ influx is the critical trigger for the induction of synaptic plasticity. We have developed a model for the induction of synaptic plasticity based on the Ca2+ influx through NMDARs by incorporating currently available data on the kinetics of NMDAR activation during synaptic activity. We find that our model accurately describes [Ca2+] when compared to previous experimental observations.

18 November 2010

Jack Mellor

Title:

A Ca2+-based computational model for NMDA receptor-dependent synaptic plasticity

Abstract:

Synaptic plasticity requires coincident activity in presynaptic and postsynaptic neurons to activate NMDA receptors (NMDARs). The resultant Ca2+ influx is the critical trigger for the induction of synaptic plasticity. We have developed a model for the induction of synaptic plasticity based on the Ca2+ influx through NMDARs by incorporating currently available data on the kinetics of NMDAR activation during synaptic activity. We find that our model accurately describes [Ca2+] when compared to previous experimental observations.

25 November 2010	Roland Baddeley (with John Fennell)
Title:	Human's representation of reward is three dimensional (reward, risk and uncertainty) and is measured by the semantic differential.
Abstract:	What characteristic(s) are represented for every object, situation and action we can make? Two traditions have provided answers to this question. Empirically, since the seminal work of Osgood (1957), factor analytic techniques have shown that over multiple different domains, cultures and languages, three underlying factors account for approximately 50% of the variance in the meaning revealed by ratings (these factors are usually labelled Evaluation, Potency and Activity). This makes this one of the most robust findings in social psychology. In contrast theoretically, whenever we wish to decide between two different options, we need a common currency to compare them. This common currency, known as utility (in economics) or reward (in psychology/biology) is often viewed as one dimensional, and decisions are proposed to maximise it. Here we propose that the factors measured by the semantic differential are actually a representation of the risk/reward structure associated with a concept. Sampling 133 million blogs entries, we found that there were extremely high correlations (R á 0.87) between: evaluation and the (log) ratio of good to bad events happening; and (R á 0.71) between potency and the (log) probability of bad things happening. Activity we propose represents the level of uncertainty due to both lack of knowledge and control. This richer three-dimensional representation of reward allows more sophisticated decisions to be made rather than simply maximising reward: specifically ones that are sensitive to risk and uncertainty. We relate this simple result to the role of neurotransmitters, cortical areas, and mining data sources for meaning.

25 November 2010

Roland Baddeley (with John Fennell)

Title:

Human's representation of reward is three dimensional (reward, risk and uncertainty) and is measured by the semantic differential.

Abstract:

What characteristic(s) are represented for every object, situation and action we can make? Two traditions have provided answers to this question. Empirically, since the seminal work of Osgood (1957), factor analytic techniques have shown that over multiple different domains, cultures and languages, three underlying factors account for approximately 50% of the variance in the meaning revealed by ratings (these factors are usually labelled Evaluation, Potency and Activity). This makes this one of the most robust findings in social psychology. In contrast theoretically, whenever we wish to decide between two different options, we need a common currency to compare them. This common currency, known as utility (in economics) or reward (in psychology/biology) is often viewed as one dimensional, and decisions are proposed to maximise it. Here we propose that the factors measured by the semantic differential are actually a representation of the risk/reward structure associated with a concept. Sampling 133 million blogs entries, we found that there were extremely high correlations (R á 0.87) between: evaluation and the (log) ratio of good to bad events happening; and (R á 0.71) between potency and the (log) probability of bad things happening. Activity we propose represents the level of uncertainty due to both lack of knowledge and control. This richer three-dimensional representation of reward allows more sophisticated decisions to be made rather than simply maximising reward: specifically ones that are sensitive to risk and uncertainty. We relate this simple result to the role of neurotransmitters, cortical areas, and mining data sources for meaning.

2 December 2010	Tijl De Bie
Title:	Could pattern mining, a branch of data mining, be of use to biological research?
Abstract:	I will outline some work I have done a few years ago on pattern mining techniques for the inferrence of regulatory networks from heterogeneous data (motifs, ChIPchip, microarray). It would be nice to discuss in general the possible use of pattern mining techniques in the reverse engineering of biological systems.

9 December 2010	Caroline Colijn
Title:	Network structures and bacterial metabolism: applications and open problems
Abstract:	Biological networks are highly structured and complex, and their structure gives rise to robust functionality at many levels. Here I will describe steady-state approaches to modelling metabolic networks, which are large biochemical networks ultimately responsible for many aspects of cellular function. Steady-state approaches are interesting because they allow modellers to work without knowledge of enzyme kinetic parameters and other quantitative features that are generally not known, especially at the genome scale. In contrast, traditional approaches such as differential equations are extremely sensitive to such quantities. I will describe some results of applying steady-state approaches to interpreting gene expression data, and then I'll describe some upcoming work, open problems and areas of promising application in this field.

9 December 2010

Caroline Colijn

Title:

Network structures and bacterial metabolism: applications and open problems

Abstract:

Biological networks are highly structured and complex, and their structure gives rise to robust functionality at many levels. Here I will describe steady-state approaches to modelling metabolic networks, which are large biochemical networks ultimately responsible for many aspects of cellular function. Steady-state approaches are interesting because they allow modellers to work without knowledge of enzyme kinetic parameters and other quantitative features that are generally not known, especially at the genome scale. In contrast, traditional approaches such as differential equations are extremely sensitive to such quantities. I will describe some results of applying steady-state approaches to interpreting gene expression data, and then I'll describe some upcoming work, open problems and areas of promising application in this field.

Forum in the previous years

2009-2010