[Comp-neuro] RODES

Alexandru Floares alexandru.floares at gmail.com
Sun Apr 25 08:31:41 CEST 2010

Dear Colleagues,

May be you are interested in RODES (Reversing Ordinary Differential
Equations Systems) - the class of algorithms I developed to automate
modeling high-throughput time course data, with ordinary differential
equations systems; they are based on computational intelligence - genetic
programming and neural network control. Please find bellow the abstract of a
book chapter and the link to full access:


Comments are welcomed.

* *

*Toward Personalized Therapy Using Artificial Intelligence Tools to
Understand and Control Drug Gene Networks*

Alexandru G. Floares

*SAIA - Solutions of Artificial Intelligence Applications;*

*IOCN - Oncological Institute Cluj-Napoca    *


The real implementation of individualized therapy and gene therapy of
multigene disorders are important goals of modern personalized medicine. A
rational foundation for this requires knowing which genes are expressed,
when, where, and to what extent. The regulation of gene expression is
achieved through complex regulatory systems - gene regulatory networks or
simply gene networks - which are networks of interactions among DNA, RNA,
proteins, and small molecules. Not only a key ingredient but a whole
dimension is missing from this view. A large variety of external molecular
species interfere with gene networks, but we will focus only on drugs, drug
discovery being one of the important routes to personalized medicine. A more
general concept of drug gene regulatory networks or simply drug gene
networks is introduced together with some mathematically definitions.
Besides the high-throughput experimental approaches, allowing to
simultaneously monitor thousands of genes or other molecular species,
mathematical modeling is essential for understanding and controlling gene
networks by drugs or gene replacements. Various formalisms, such as Bayesian
networks, Boolean networks, differential equation models, qualitative
differential equations, stochastic equations, and rule-based systems, have
been used. The ordinary differential equations approach tries to elucidate a
deeper understanding of the exact nature of the regulatory circuits and
their regulation mechanisms, but is also difficult. There is a need for
algorithms to automatically infer such models from high-throughput
time-series data, and artificial intelligence is better suited than
conventional modeling. We proposed a reverse engineering algorithm for drug
gene networks, based on artificial intelligence methods: neural networks for
identification and control, and genetic programming for symbolic regression.
It takes as inputs high-throughput (e.g., microarray) time-series data and
automatically infer an accurate ordinary differential equations model,
revealing the networks structure and parameter and giving insights into the
molecular mechanisms involved. RODES, from reversing ordinary differential
equations systems, decouples the systems of differential equations, reducing
the problem to that of revere engineering individual algebraic equations.
Usually, due to various experimental constraints, essential information is
missing from data, and even the most powerful artificial intelligence
techniques are not creating information but just extracting it from data. In
the present context, not all variables or time-series are simultaneously
measured, as it is required to reconstruct the drug gene networks, as
systems of ordinary differential equations. One of the unique features of
RODES is its ability to deal with the common but challenging situations of
information missing from data. Thinking in a systemic way one can conjecture
that, due to the interactions in these networks, information must be
implicitly present in the data. Most if not all data mining techniques are
dealing exclusively with explicit information extraction from data.
Therefore we used some ideas from control theory, choosing a simple but
powerful technique, feedback linearization. To automate the algorithm the
neural networks counterpart of the conventional method was used. Applied to
drug gene networks this algorithm enable and automate the  reconstruction of
the time-series of the transcription factors, microRNA, or drug related
compounds which are usually missing in microarray experiments. RODES is also
able to incorporate common a priori knowledge. To our knowledge, this is the
first realistic reverse engineering algorithm, based on genetic programming
and neural networks, applicable to large gene networks.

Alexandru Floares, MD, PhD

Head of
Artificial Intelligence Department
Cancer Institute Cluj-Napoca
400015, Str. Republicii, Nr. 34-36,
Cluj-Napoca, Romania
Email: alexandru.floares at iocn.ro

President of
SAIA Group
400310 Str. Al. Vlahuta, Bl. Lama C, Ap. 45,
Cluj-Napoca, Romania
Email: alexandru.floares at saia.ro
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://www.neuroinf.org/pipermail/comp-neuro/attachments/20100425/6eea98d5/attachment.html

More information about the Comp-neuro mailing list