Hello, I’m Yoshiyuki Asai, a
group leader of open biology unit in
Okinawa Institute of
Science and Technology, Japan.
In this lecture I want to introduce a versatile platform for multilevel modeling of physiological systems.
In recent biological and physiological
research, computable mathematical models have
become increasingly important for integrating
the huge amount of knowledge and data obtained
from experiments and simulations, and for applying simulation results to medicine.
One of dream in such integrative
physiology is to cause a paradigm shift from
empirical medicine to predictive medicine.
To promote effective collaborations
to build large-scale models,
it is also important to consolidate
fundamental tools to support such activities.
Model sharing and model reuse, which are
crucial for the above-mentioned
multidisciplinary collaborations, must be
encouraged by using such tools.
There have been several pioneering efforts to develop technologies in that direction
such as SBML, CellML and PHML,
These are XML-based descriptive
to describe the dynamics of biological
and physiological systems.
The main purpose of the
development of these languages was
to establish a common
to enhance the exchange of
models among collaborators.
Research in this kind of direction is
sometimes referred to as Physiome,
which is one of omics such
as genome and proteome.
It is considered that the first step
what must be done in Physiome
is to develop comprehensive
methods for acquisition and databasing
of very large sets of information on
all aspects of biology
to share and reuse the data among researchers.
Then construct descriptive and quantitative
models and organize collaborations.
The target to understand and build a
database is functions of living organisms.
A physiological function is a change of
the state of living organisms in time,
that is dynamics of physiological state.
It is impossible to store the dynamics in a database.
Of course it is possible to store
experimental time series data in a database,
and it is important for physiome
as the first step indeed.
But such time series represents only one
case of the dynamics of physiological function
which happened under the experimental condition.
Here physiologically plausible
mathematical models play important roles.
Since the models are symbolically
written and so can be stored in databases.
Besides, anytime we can take out
the dynamics of the modeled function
by numerical integration,
Even more importantly,
by changing parameters,
we can observe dynamics at the different
conditions rather easily.
Since the modeling target is ranging
from molecular level up to individual level,
variety of modeling techniques
must be incorporated
to build such multilevel models.
For example, agent based
systems are sometimes used for
molecular dynamics simulation.
Ordinary differential equations cover
a wide range of dynamics such as
membrane potential and walking motion.
Of course partial differential equations
and algebraic equations are also used.
Since the modeling targets are
multilevel to consider not only
intra-level principles but also
these techniques are often
used together in one model.
The slide shows the entire scheme of
our system including applications and databases.
To support users to build such
models of multilevel physiological systems,
we have been developing PhysioDesigner
on which users can make their own models.
We also developed databases
of models, and morphological
and time series data.
Models developed on PhysioDesigner
are written in PHML format,
which is an XML based
model descriptive specification.
And finally, simulations of those
models are performed by Flint,
a simulator supporting
SBML as well as PHML.
What is the benefit of using
Firstly it is possible to describe a
model with the additional information,
such as, article introducing
the model, model creator,
description of the model, and so on.
Then the model would be a quit self-consistent.
For example, when other
researchers read the model,
he or she can understand
immediately what the model is,
to who should have a
contact, and so on.
The second. On this platform,
modeling and performing a simulation
are completely separated.
Modeling is done on PhysioDesigner,
and simulation is done on Flint.
This means that model creators
can concentrate on modeling and
logics to be modeled, and do not
need to be bothered by programming,
especially of implementation of
This aspect becomes very
important when one thinks
of parallel computing. Usually it
requires high programming skill
to implement. Nowadays
models tend to be
progressively larger, parallel
becomes more important.
The third point is about
openness of the models.
Sharing and reuse of models
is crucial point to build a
large scale multi level model.
Of course people can share
source codes of models
even they use computer languages
such as Java, C++ to build
But usually it is not easy to
read source codes
written by others. Moreover
it is difficult to merge
them into one model.
PHML is a text format with
tags, including a lot of
additional information, which
makes the model easier to read.
Besides if models are opened,
it is very easy to
reproduce and validate the
published simulation results.
This enhances soundness
of simulation research.