Maria Schilstra's ongoing projects

NetBuilder and NetBuilder-prime ('Apostrophe')

Genetic regulatory networks (GRNs) contain and dispense the information necessary for embryonic development, the information required for organisms to adapt to their environment, and for numerous other processes. GRNs are formed by genes that contain the information required to form transcription factors, proteins that can initiate or stop the formation of other protein molecules. Stimulation or suppression of protein formation is very specific: a certain type of transcription factor will directly affect transcription of, at most, a few genes, and often only in a particular combination with transcription factor molecules of different types. NetBuilder was our first attempt at a graphical tool for building logical representations of multicellular models of GRNs. You can find documentation, examples of GRNs simulated with NetBuilder, and instructions on how to download and install the last version of the old program from its legacy website.

The NetBuilder project is continued under the name Apostrophe (after the Frank Zappa song), and the completely overhauled version of the tool is called NetBuilder' (NetBuilder-apostrophe, or NetBuilder-prime). This is its Source-forge site. More information on NetBuilder' can be found on this website by clicking on the picture or by following this link.

Applying biological control principles to autonomous decision-making in manufacturing

We are collaborating with the Lean Engineering Research Group at the Centre for Manufacturing at De Montfort University in Leicester to find out if biological control principles can be applied to improve autonomous decision making in industrial operational systems. The objectives of of this EPSRC-funded project are summarized below.
There are fundamental differences between the Modelling and Simulation (M&S) requirements of biological control and regulation networks and those of Industrial Operational Systems (IOSs). One fundamental difference is the way M&S approach the design and planning tasks undertaken by both systems. Within a biological cell these tasks are an integral part of the 'control' processes whereas in IOSs they are normally external tasks which are undertaken prior to the operation of such systems and which act as constraints on their control systems. The first aim of this project is to examine, by means of discrete event simulation (DES), the feasibility of using biological control principles to significantly increase levels of autonomous decision-making within IOSs, through incorporating their design, planning and scheduling activities within their control processes. The chance of discovering fundamental autonomous control principles represents an exciting prospect.
Systems biologists make use of a range of modelling and simulation techniques in their attempts to represent and analyse biological control and regulation networks. There is much discussion in the research literature as to which techniques are most appropriate for modelling purposes. The generally accepted, thermodynamics-based, modelling context yields a low-level non-hierarchical description of the system as a series of chemical reactions. However, without an interpretation of the system in terms of hierarchy and function, it is difficult to grasp the significance of any simulation results, and to evaluate the response of a system to external signals. A platform that considers the functional hierarchy of a compound system could significantly facilitate the analysis of the system's responses to external signals, changed conditions, and other perturbations. The second aim of this project is to examine the feasibility of using DES to provide a platform for integrating low-level biochemical reactions with their higher level function and control processes.
This work has potential for significantly improving biological M&S methods. The third aim is to take the opportunity provided by this project, through its high level of cross-disciplinary activities, to examine the relevance of potentially radical perspectives, i.e. that of industrial and manufacturing systems, that could aid the generation of novel and potentially ground-breaking hypotheses on the control of intracellular manufacturing processes.

An interim report by Mark McAuley, who carried out our side of the research, is found by following this link.

CellML to SBML conversion

It all started when I was putting new models of biochemical reaction networks into the old SBML Models repository - now replaced by the BioModels Database. Because the CellML Team had already done all the work (searching the literature, interpreting papers, etc.), the CellML Models Repository was full of of highly delectable models. A CellML to SBML converter was called for. I produced one, in XSLT, which I used to convert the whole CellML database (over 300 models) into SBML, with a success rate of about 90%. The remaining 10% was mostly due to minor incompatibilities between CellML and SBML, and the score could be improved to about 96% by manually making some inconsequential alterations to the CellML models. You can download the XSLT stylesheets, and try for yourself. However, even though the converter works well in relatively straightforward cases, it is far from perfect, and not quite general either. Therefore, the capable Lu Li and Nicolas Le Novère at the EMBL-EBI (home of the BioModels Database) have taken over further development of the converter. A specification of the minimal requirements for such a converter, and a description of the first attempt has been published here (Bioinformatics)

BoB (stochastic simulation of a process with non-Markov dynamics)

Myosin-V is a molecular motor that 'walks' along the ubiquitous actin tracks in cells, and drags membraneous vesicles (which can be filled with, for instance, neurotransmitters) to their destination (such as synapses). ATP hydrolysis fuels the motion. The actual stepping, where one leg is swung and put in front of - or occasionally behind - the other, is very fast. However, the intervals between the steps, where both feet are firmly stuck to the track, are much longer, and distributed randomly - at least when observed in vitro. In vivo, where there is no feedback mechanisms that keeps the force on the motor constant, the cargo is connected via an elastic tether, which stretches - and creates extra pull - every time the motor steps forward. The bulky cargo has to move through the highly viscous cytosol, and follows much more slowly. Thus, the pull on the motor decreases in the interval between steps. Importantly, the stepping rate of the motor is dependent on the force it experiences. To simulate this process properly, it is necessary to incorporate the fact that the stepping probability may change a lot in the interval between two steps, and that is what I did when I wrote BoB. The results of this exercise are published here (J. Royal Soc. Interface)