Information Processign in Cells and Tissues

Artificial gene regulatory networks are computational models which draw inspiration from real world networks of biological gene regulation. Since their inception they have been used to infer knowledge about gene regulation and as methods of computation. These computational models have been shown to possess properties typically found in the biological world such as robustness and self organisation. Recently, it has become apparent that epigenetic mechanisms play an important role in gene regulation. This paper introduces a new model, the Artificial Epigenetic Regulatory Network (AERN) which builds upon existing models by adding an epigenetic control layer. The results demonstrate that the AERNs are more adept at controlling multiple opposing trajectories within Chirikov’s standard map, suggesting that AERNs are an interesting area for further investigation.

We describe a simulation of multi-genic heterosis using Boolean gene regulatory networks. Hybrid vigor, or heterosis, is the phenomenon whereby the offspring of crosses from separate populations often perform better than inbreds with respect to growth rate, fertility and disease resistance. Because of its great economic importance, the genetic and molecular basis of heterosis has been subject of many scientific studies. However, attempts to model the phenomenon from a systems biology point of view have been quite abstract.

Our model allows the generation, evolution, homologous recombination and hybridisation of networks which display the properties of gene regulatory networks observed in biology in a simulated environment. We can thus test the current hypotheses about heterosis and investigate which factors affect it.

We compare the discrete asynchronous logical modeling formalism for gene regulatory networks due to R. Thomas with piecewise-affine differential equation models. We show that although the two approaches are based on equivalent information, the resulting qualitative dynamics are different.

Advanced mathematical methods and computational tools are required to properly understand the behavior of large and complex regulatory networks that control cellular processes. Since available data are predominantly qualitative or semi-quantitative, discrete (logical) modeling approaches are increasingly used to model these networks. Here, relying on the multilevel logical formalism developed by R. Thomas

et al

. [7,9,8], we propose a computational approach enabling (i) to check the existence of at least one consistent model, given partial data on the regulatory structure and dynamical properties, and (ii) to infer properties common to all consistent models. Such properties represent non trivial deductions and could be used by the biologist to design new experiments. Rather than focusing on a single plausible solution,

i.e.

a model fully defined, we consider the whole class of models consistent with the available data and some economy criteria, from which we deduce shared properties. We use constraint programming to represent this class of models as the set of all solutions of a set of constraints [3]. For the sake of efficiency, we have developed a framework, called SysBiOX, enabling (i) the integration of partial gene interaction and expression data into constraints and (ii) the resolution of these constraints in order to infer properties about the structure or the behaviors of the gene network. SysBiOX is implemented in ASP (Answer Set Programming) using Clingo [4].

CRISPR is a recently discovered adaptive prokaryotic immune system. A crucial step in CRISPR defense mechanism is transcription of CRISPR cassette, which is followed by processing of the resulting long transcript (pre-crRNA) into small RNA molecules (crRNA) that recognize invading viruses. We model CRISPR transcript processing, and show that the system functions as a strong amplifier, which can rapidly generate a large number of crRNAs from only few pre-crRNA molecules. Based on this analysis, we propose a synthetic gene circuit that can produce a large number of desired molecules from a potentially toxic substrate.

Site specific transcription factors (TF) are proteins that orchestrate transcription by binding to specific target sites on the DNA. This binding can be both sequence- and conformation-specific. However, also non-specific binding with lower affinity can be observed [3]. The number of specific target sites is significantly smaller compared to the number of non-specific sites and, consequently, TF molecules bind, in a first instance, non-specifically to the DNA. Once bound to the DNA the TF molecules perform an one dimensional random walk on the DNA until they either find a target site or unbind from the DNA template. In particular, during the one dimensional random walk on the DNA, a molecule will perform one of the three types of movements: (

i

) sliding , (

ii

) hopping and (

iii

) jumping [6]. This combination of one and three dimensional diffusion is called

facilitated diffusion

and it is hypothesised that this speeds up the search process [3,2,5].

Artificial Signalling Networks (ASNs) are computational models inspired by cellular signalling processes that interpret environmental information. This paper introduces an ASN-based approach to controlling chaotic dynamics in discrete dynamical systems, which are representative of complex behaviours which occur in the real world. Considering the main biological interpretations of signalling pathways, two ASN models are developed. They highlight how pathways’ complex behavioural dynamics can be captured and represented within evolutionary algorithms. In addition, the regulatory capacity of the major regulatory functions within living organisms is also explored. The results highlight the importance of the representation to model signalling pathway behaviours and reveal that the inclusion of crosstalk positively affects the performance of the model.

Membrane receptors allow the cell to respond to changes in the composition of its external medium. The ligand-receptor interaction is the core of the signalling process and may be greatly influenced by the spatial configuration of receptors. As growing pieces of evidence suggest that receptors are not homogeneously spread on the cell surface, but tend to form clusters, we propose to investigate the implication of receptor clustering on ligand binding kinetics using a computational individual-based model. The model simulates the activation of receptors distributed in clusters or uniformly spread. The tracking of binding events allows the analysis of the effect of receptor clustering through the autocorrelation of the receptor activation signal and the empirical time distributions of binding events, which are still unreachable with in vitro or in vivo experiments. Results show that the apparent affinity of clustered receptors is decreased. Additionally, receptor occupation becomes spatially and temporally correlated, as clustering creates platforms of coherently activated receptors. Changes in the spatial characteristics of a signalling system at the microscopic scale globally affect its function in time and space.

To be able to understand the mechanisms of action of drugs, predict their efficacy, and anticipate their potential side-effects is important during drug development. In diseases where the genetic background of patients modulates treatment response, it might allow personalizing the therapy.

Substantial progress in proteomic technologies[1] have made it possible to develop chemical proteomics methods, where the protein targets of a drug are affinity-purified and identified by mass spectrometry[2, 3]. Compound-protein interactions are measured in a biological context as opposed to in

vitro

binding assays. That is, drugprotein interactions can not only be determined proteome-wide, but also in a tissue- or cell type-dependent manner.

The reversible postranslational modification of proteins is a ubiquitous feature of cellular signal transduction networks. In these systems, signalling is typically seen as resulting from the interaction between an active enzyme and a downstream unmodified substrate. However, it is known that in some cases the inactive form of an enzyme is also capable of binding the unmodified substrate, and that in other cases the active enzyme is capable of binding modified substrate. In this paper, we analyse the behaviour of a two-stage enzymatic cascade in which these additional protein-protein interactions are possible. Without the additional interactions, the model produces the standard ultrasensitive switch-like behaviour. We find that inactive enzyme binding to unmodified substrate increases the ultrasensitivity of this switch, while active enzyme binding to modified substrate results in the switch becoming biphasic. These results indicate how important the rules governing the occurrence of protein-protein interactions can be in determining the signalling behaviour of a pathway, even when particular protein-protein interactions have no clear functional role.

The three-dimensional structure of a protein can be described as a graph where nodes represent residues and interactions between them are edges. We have constructed protein contact networks at different length-scales for different interaction strength cutoffs. The largest connected component of short-range networks exhibit a highly cooperative transition, while long- and all-range networks (more similar to each other), have less cooperativity. The hydrophobic subnetworks in all- and long-range networks have similar phase transition behaviours while hydrophilic and charged networks don’t. Hydrophobic subclusters in long- and all-range networks exhibit higher occurrence of assortativity and hence higher communication ability in transmitting information within a protein. The highly cliquish hydrophobic nodes in long- and short-range networks play a significant role in bridging and stabilizing distantly placed residues during protein folding. We have also observed a significant dominance of charged residues cliques in short-range networks.

Differences in tissues arising from a single organism are attributable, at least partially, to differing metabolic regimes. A highly topical instance of this is the Warburg effect in tumour development, whereby malignant tissue exhibits greatly altered metabolism compared to healthy tissue. To this end, we consider the emergent properties of two metabolomic datasets from a human glioma cell line (U87) and a human mesenchymal stem cell line (hMSC). Using a random matrix theory (RMT) approach, U87 is found to have a modular structure, whereas hMSC does not. The datasets are then compared using between groups comparison of principal components, and finally, a group of metabolites is found that remains highly correlated in both conditions.

We demonstrate the power of evolutionary robotics (ER) by comparing to a more traditional approach its performance and cost on the task of simulated robot locomotion. A novel quadruped robot is presented, the legs of which – each having three non-coplanar degrees of freedom – are very maneuverable. Using a simplistic control architecture and a physics simulation of the robot, gaits are designed both by hand and using a highly parallel evolutionary algorithm (EA). It is found that the EA produces, in a small fraction of the time that takes to design by hand, gaits that travel at nearly twice the speed of the hand-designed one.

The nematode

Caenorhabditis elegans

has been established as a modeling organism in biomedical research for several decades. Its hermaphrodite germ line encompasses key developmental concepts like stem cell differentiation and apoptosis; therefore it provides a good model system for these basic concepts. Here, we have extended and refined our previous computational model, which encompasses developmental landmarks and the resulting movement of germ cells along the gonadal tube. We have used the molecular dynamics (MD) framework to model the physical movement of cells due to the force arising from cell divisions. The model simulation was calibrated with experimental time and it is in accordance with experimental observations. In addition, the model provides means for testing hypotheses regarding the behaviour of mutated germ lines and the potential mechanisms causing physiological apoptosis, which are difficult to assess experimentally.

The tissues of multicellular organisms are made of differentiated cells arranged in organized patterns. This organization emerges during development from the coupling of dynamic intra- and intercellular regulatory networks. This work applies the methods of information theory to understand how regulatory network structure within and between cells relates to the complexity of spatial patterns that emerge as a consequence of network operation. A computational study was performed in which undifferentiated cells were arranged in a two dimensional lattice, with gene expression in each cell regulated by an identical intracellular randomly generated Boolean network. Cell-cell contact signalling between embryonic cells is modeled as coupling among intracellular networks so that gene expression in one cell can influence the expression of genes in adjacent cells. In this system, the initially identical cells differentiate and form patterns of different cell types. The complexity of network structure, temporal dynamics and spatial organization is quantified through the Kolmogorov-based measures of normalized compression distance and set complexity. Results over sets of random networks from ordered, critical and chaotic domains demonstrate that: (1) Ordered and critical intracellular networks tend to create the most complex intercellular communication networks and the most information-dense patterns; (2) signalling configurations where cell-to-cell communication is non-directional mostly produce simple patterns irrespective of the internal network domain; and (3) directional signalling configurations, similar to those that function in planar cell polarity, produce the most complex patterns when the intracellular networks are non-chaotic.

The task of the vasopressin system is homeostasis, a type of process which is fundamental to the brain’s regulation of the body, exists in many different systems, and is vital to health and survival. Many illnesses are related to the dysfunction of homeostatic systems, including high blood pressure, obesity and diabetes. Beyond the vasopressin system’s own importance, in regulating osmotic pressure, it presents an accessible model where we can learn how the features of homeostatic systems generally relate to their function, and potentially develop treatments. The vasopressin system is an important model system in neuroscience because it presents an accessible system in which to investigate the function and importance of, for example, dendritic release and burst firing, both of which are found in many systems of the brain. We have only recently begun to understand the contribution of dendritic release to neuronal function and information processing. Burst firing has most commonly been associated with rhythm generation; in this system it clearly plays a different role, still to be understood fully.

Calcium/calmodulin-dependent protein kinase II (CaMKII) is involved in the induction of many forms of synaptic plasticity in the brain. Experimental and computational studies have shown that CaMKII is sensitive to the frequency of oscillatory Ca

2 + 

signals. Here we demonstrate that in a simple, commonly used kinetic model of CaMKII phosphorylation, the overall phosphorylation rate under sustained application of Ca

4

 − CaM pulses ultimately depends on the average (‘effective’) concentration of Ca

4

 − CaM in the system, rather than on the pulse frequency itself. As a corollary, equal phosphorylation levels are achieved in response to pulsed and constant applications of equal effective concentrations of Ca

4

 − CaM.

In real neuronal networks it is known that neurons are either excitatory or inhibitory. However, it is not known whether all synapses within the subpopulations are plastic. It is interesting to investigate the implications these constraints may have on functionality. Here we investigate highly simplified models of associative memory with a variety of allowed synaptic plasticity regimes. We show that the allowed synaptic plasticity does indeed have a large effect on the performance of the network and that some regimes are much better than others.

Diffusive Computation is a method of using diffusing particles as a representation of data. The work presented attempts to show that through simulating spiking neurons, diffusive computation has at least the same computational power as spiking neural networks. We demonstrate (by simulation) that wavefronts in a Reaction-Diffusion system have a cumulative effect on concentration of reaction components when they arrive at the same point in the reactor, and that a catalyst-free region acts as a threshold on the initiation of an outgoing wave. Spiking neuron models can be mapped onto this system, and therefore RD systems can be used for computation using the same models as are applied to spiking neurons.

Investigation of a biological domain through simulation can naturally lead to the desire to extend the simulation as new areas of the domain are explored. Such extension may entail the incorporation of additional cell types, molecules or entire molecular pathways. The addition of these extensions can have a profound influence on simulation behaviour, and where the biological domain is not well characterised, a structured development methodology must be employed to ensure that the extended simulation is well aligned with its predecessor. The paper presents such a methodology, relying on iterated development and sensitivity analysis, by extending an existing simulation of Experimental Autoimmune Encephalomyelitis (EAE), a disease model for Multiple Sclerosis, via inclusion of an additional regulatory pathway. We reflect on the implications of extensions which alter simulation behaviour on pre-extension results.

The mechanisms leading to the initiation of normal, premature or dysfunctional human labour are poorly understood, as animal models are inappropriate, and experimental studies are limited. Computational modelling provides a means of linking non-invasive clinical data with the results of

in vitro

cell and tissue physiology. Nonlinear wave processes – propagation in an excitable medium – provides a quantitatively testable description of mechanisms of premature and full term labour, and a view of changes in uterine electrophysiology during gestation as a trajectory in excitation and intercellular coupling parameter space. Propagation phenomena can account for both premature and full term labour.

Artificial biochemical networks (ABNs) are a class of computational automata whose architectures are motivated by the organisation of genetic and metabolic networks. In this work, we investigate whether evolved ABNs can carry out classification when stimulated with time series data collected from human subjects with and without Parkinson’s disease. The evolved ABNs have accuracies in the region of 80-90%, significantly higher than the diagnostic accuracies typically found in initial clinical diagnosis. We also show that relatively simple ABNs, comprising only a small number of discrete maps, are able to recognise the abnormal patterns of motor function associated with Parkinson’s disease.

Systems biology is very much concerned with gaining an overview of what is happening in complex systems, such as in biomedical data sets, for which we need good global visualization tools. This research uses a method based on information distance geometry to create visualizations analogous to heat-maps of prognostic and diagnostic variables. It illustrates the advantages of an informationally self-structuring approach to the understanding of biomedical data.

Over the course of billions of years, under evolutionary pressure, Nature has evolved solutions to various problems. As our ability to understand the biological mechanisms that are intrinsic in these solutions continues to improve, we have the opportunity to apply this knowledge when solving our challenging problems, in fields such as medicine and the environment. This paper discusses an approach, in which biological systems are investigated as information processing systems, and the understanding of how these systems process information is then applied to engineering systems. Two examples are presented. The first one discusses how the heart’s fault-tolerant information processing can be implemented in an electronic system. The second example discusses a cellular biochemical reaction network and how its property of robustness can be implemented in a chemical system. Finally, three different applications, in which this approach is already being applied with promising results, are briefly reviewed.

An organism’s behaviour can be categorised as being either predatory or anti-prey. Predatory behaviours are behaviours that try to improve the life of an organism. Anti-prey behaviours are those that attempt to prevent death. Regulation between these two opposing behaviours is necessary to ensure survivability—and gene regulatory networks and metabolic networks are the mechanisms that provide this regulation. We know that such regulatory behaviour is encoded in an organism’s genes. The question is, how is it encoded? The understanding of this encoding can help with the development of an artificial organism, for example an autonomous robotic system; whereby the robot will have the ability to autonomously regulate the switching between the opposing behaviours using this encoded mechanism, in order to ensure its sustainable and continuous system operations. This paper aims to look into the properties of an artificial bio-chemical network consisting of a genetic regulatory network and a metabolic network that can provide these capabilities.

Examination of the scalar properties of living organisms and the electronic configuration of crystalline structures suggests that similar modeling may be used for both. This paper comments on individual and common properties of the two systems and draws a comparison between them. Both exhibit multiple scales and a global ‘overview’ of their scalar properties. We conclude that the analogy may provide a fruitful route towards modeling living organisms.

We analyse two very simple Petri nets inspired by the Oregonator model of the Belousov-Zhabotinsky reaction using our stochastic Petri net simulator. We then perform the Krohn-Rhodes holonomy decomposition of the automata derived from the Petri nets. The simplest case shows that the automaton can be expressed as a cascade of permutation-reset cyclic groups, with only 2 out of the 12 levels having only trivial permutations. The second case leads to a 35-level decomposition with 5 different simple non-abelian groups (SNAGs), the largest of which is

A

9

. Although the precise computational significance of these algebraic structures is not clear, the results suggest a correspondence between simple oscillations and cyclic groups, and the presence of SNAGs indicates that even extremely simple chemical systems may contain functionally complete algebras.

It is known that much of the genetic change underlying morphological evolution takes place in

cis

-regulatory regions, rather than in the coding regions of genes. Identifying these sites in a genome is a non-trivial problem. Experimental methods for finding binding sites exist with some limitations regarding their applicability, accuracy, availability or cost. On the other hand predicting algorithms perform rather poorly. The aim of this research is to develop and improve computational approaches for the prediction of transcription factor binding sites (TFBSs) by integrating the results of computational algorithms and other sources of complementary biological evidence, with particular emphasis on the use of the Support Vector Machine (SVM). Data from two organisms, yeast and mouse, were used in this study. The initial results were not particularly encouraging, as still giving predictions of low quality. However, when the vectors labelled as non-binding sites in the training set were replaced by randomised training vectors, a significant improvement in performance was observed. This gave substantial improvement over the yeast genome and even greater improvement for the mouse data. In fact the resulting classifier was finding over 80% of the binding sites in the test set and moreover 80% of the predictions were correct.

After the sequencing of whole genomes and the identification of the genes contained in them, one of the main challenges remaining is to understand the mechanisms that regulate the expression of genes within the genome in order to gain knowledge about structural, biochemical, physiological and behavioral characteristics of organisms. Some of these mechanisms are controlled by so-called Genetic Regulatory Networks (GRNs). Boolean networks can help model biological GRNs. In this paper, a genetic algorithm is used to make inferences in Boolean networks, in combination with the Quine-McCluskey algorithm, when not all the output states of the genes have been determined. This lack of information could be treated as “don’t care” states. Genetic algorithms are useful in multi-objective optimization problems, such as minimization of Gene Regulatory Functions, where it is important not only to have the smallest quantity of disjunctions, but also the smallest quantity of genes involved in the regulation.

Suppose that we wish to know a group of proteins responsible for a certain cellular biological process. Here we propose to infer such a protein complex from a protein-protein interaction network by using a class of algorithm, which has originally been developed to achieve web page ranking that reflects user’s personal interest or context. The inference of proteins responsible for a given biological process, namely, personalized ranking of proteins is whereby performed in analogy with personalized ranking of web pages. Searching for the best approach to personalized protein ranking, we carry out a series of experiments to compare the performance between two major personalized ranking methods: the personalized PageRank algorithm and the continuous-attractor dynamics algorithm, both applied to a yeast protein-protein interaction network. Results of these comparison experiments suggest that the continuous-attractor dynamics algorithm is the most efficient for personalized protein ranking.

The search of a multiple sequence alignment (MSA) is a well-known problem in bioinformatics that consists in finding a sequence alignment of three or more biological sequences. In this paper, we propose a parallel iterative algorithm for the global alignment of multiple biological sequences. In this algorithm, a number of processes work independently at the same time searching for the best MSA of a set of sequences. It uses a Longest Common Subsequence (LCS) technique in order to generate a first MSA. An iterative process improves the MSA by applying a number of operators that have been implemented to produce more accurate alignments. Simulations were made using sequences from the UniProKB protein database. A preliminary performance analysis and comparison with several common methods for MSA shows promising results. The implementation was developed on a cluster platform through the use of the standard Message Passing Interface (MPI) library.