source: liacs/pnbm/project/report.tex@ 273

%
% $Id: report.tex 571 2008-04-20 17:31:04Z rick $
%

\documentclass[12pt,a4paper]{article}

\frenchspacing
\usepackage[english]{babel}
\selectlanguage{english}
\usepackage{graphicx}
\usepackage{url}
\usepackage{multicol}
\usepackage{fancybox}
\usepackage{amssymb,amsmath}

\title{Modeling planar signalling in AP axis development in \emph{Xenopus laevis}\\
\large{using Petri Nets in Higher Level Developmental Biology}}
\author{Rick van der Zwet, Tiago Borges Coelho \\
  \texttt{<hvdzwet@liacs.nl>,<borges.coelho@gmail.com>}\\ 
  LIACS - Leiden University, The Netherlands}
\date{\today}

\begin{document}
\maketitle
\section{Abstract}
Planar signaling is a process that is part of the development of the AP axis in
\emph{Xenopus laevis} \cite{Bertens09}, in which cells accumulate proteins
based on the saturation of nearby cells. If one cell produces n amount of
proteins, it will initiate a transferring cascade to cells in the vicinity.
This dissemination of proteins will eventually cease, considering that n is a
finite variable. There is a gradation in the amount of proteins transferred,
meaning that neighboring cells get n/2 the amount of proteins of the most
saturated cell.

We are going to model this into Petri-Nets, since it supports a mathematical
modeling language as well as a graphical interactive representation, which is
well suited for this purpose, as we could model the process based on algorithms
and also used it for automated model tracking and analysis, while having a
visual representation of the process.

\section{Method}
Our approach to modeling Planar signaling is firstly construct a biological
model of the process, one that illustrates the phenomenon taking place (shown
at figure~\ref{fig:ab-model}). Next we decided to create a conceptual model,
one that abstracts from the biological model and contains some mathematical
equations that describe the processes occurring in nature.

\begin{figure}[htp]
\centering
\caption{Conceptual abstract model}
\includegraphics[width=\textwidth]{frog-model.eps}
\label{fig:ab-model}
\end{figure}

Finally we are going to construct a Petri-Net model using the software
\emph{CPNTools} \footnote{http://wiki.daimi.au.dk/cpntools/cpntools.wiki}.

\section{Modeling}
\begin{figure}[htp]
\centering
\caption{Planar signaling model}
\includegraphics[width=100mm]{planar-signaling-model.eps}
\label{fig:model}
\end{figure}

For the conceptual model (Fig.~\ref{fig:model}) we start with a building block
that is an abstraction of a cell (figure: circle), which can then be coupled to
other cells (figure: arrows). The abstraction contains two different token
types. For one the proteins are modeled through red tokens, and secondly the
proteins generate a second process, the level of posterisation (blue tokens)
which is also required in the model. We assume a 1:1 mapping between the amount
of proteins and the posterisation, taking into consideration that when an
\texttt{INITIAL} protein is ’used’ (e.g. has a posterisation counterpart) in
this process is called \texttt{ACTIVATED}. We assume that the “proteins to
posterisation” process is taking place at the same time as the proteins
distribution. This is represented in a special format, represented by the
square B in the figure. It tries to match the posterisation to the same level
as the proteins present. The moment the protein level lowers, the posterisation
will remain the same. In pseudo-code:
\begin{verbatim}
if numPos < numProteins then
  numPos = numPos + 1
endif
\end{verbatim}

\texttt{numProteins} is the proteins available and \texttt{numPos} is the
posterisation present.

The connectors between the cells (the membranes) have a special property. One
can see them as pressure valves (like figure~\ref{fig:pressure}) that close
when the ’volume’ in the containers (cells) complies with the following
property $A/2 < B$, or siphons when the property before is not achieved,
passing volume from $A$ to $B$ at a certain rate (\texttt{flowSpeed}). This
rate could depend on the difference, actual value present or something else.
Please keep in mind that negative values could sometimes appear, hence there is
a need to check whether the source is bigger or equal than the
\texttt{flowSpeed}. 

\begin{figure}[htp]
\centering
\caption{Pressure valve example}
\includegraphics[height=60mm]{pressure-valve.eps}
\label{fig:pressure}
\end{figure}

For this case there exists no standard Petri-Net ’component’, hence this
requires the creation of a new property, which is described in pseudo-code
below:
\begin{verbatim}
flowSpeed = n
if A > 2 * B and A => flowSpeed then
  A = A - flowSpeed
  B = B + flowSpeed
else if B > 2 * A and B => flowSpeed then
  B = B - flowSpeed
  A = A + flowSpeed
endif
\end{verbatim}

Planar signaling could theoretically start in every cell, by inserting some
amount of proteins. In our model this is represented as $n$ \texttt{INITIAL}
tokens being put in a cell chosen at random.


\section{CPNTools implementation}
CPNTools has some shortcomings when it comes to modeling higher level
developmental biology. One is the shortcoming of ’balancing’. It does not allow
the reading of how many tokens are present in a certain state and base action
upon them.

As workaround for this (see Fig~\ref{fig:CPNplanar}) we used a ’dump’ gradation
function.  In our case it simply takes 3 tokens and pushes 1 forward while
converting 2 directly to \texttt{ACTIVATED}. This does not take into
consideration if some external source changes the amount further up. 

Secondly it misses the possibility to easily randomize the initialization. As a
quick fix we ’hacked’ it to choose between starting at the head or the tail. In
this implementation the protein to gradient process is taking place at cell $A$
at the same time that the proteins get transferred from cell $A$ to $B$. It
should be noted that the model avoids the notion of timed firing sequences;
meaning that the firing sequences will not occur at pre-determined times. This
could be changed in the future to ‘trigger’ a timed activation of the
\texttt{INITIAL} to \texttt{ACTIVATED} process as modeled in
figure~\ref{fig:model}.

\begin{figure}[htp]
\centering
\caption{CPNTools implementation}
\advance\leftskip-2cm
\advance\rightskip+2cm
\includegraphics[width=1.3\textwidth]{planar-signaling.eps}
\label{fig:CPNplanar}
\end{figure}

\section{Conclusion}
Using Petri-Nets for modeling biology processes is a powerful framework, one
which could be well expandable. The proof of concept implementation and
visualization however is lacking. \emph{CPNTools} does not currently provide a
powerful enough tool-set for the modeling purposes. This could be improved with
the support of a programming language that can produce algorithms that can
replace the mathematical functions of arcs.



\bibliographystyle{amsalpha}
\begin{thebibliography}{10}
\bibitem[Bertens09]{Bertens09}Laura M.F. Bertens et al., Using Petri Nets in Higher
Level Developmental Biology: A case study on the AP axis development in Xenopus
laevis Extended Abstract, 2009
\end{thebibliography}
\end{document}