http://cbi-society.org/cbi/taikai/taikai11/index_e.html Edgar Wingender: "Efficient large-scale screening of chemical structures for potential biological activities" "The geneXplain platform: from functional genome analysis to biological process simulation" "One out of 24 million: How to computationally identify a new candidate drug against breast cancer" geneXplain GmbH

1. Efficient large-scale screening of chemical structures for potential biological activities
Making use of a very large data basis that has been gathered over about 40 years, the software PASS
can efficiently predict biological activities of chemical substances. The acronym PASS stands for
Prediction of Activity Spectra for Substances. This allows the evaluation of the biological activity
profiles for compounds even prior to their chemical synthesis and biological testing. Moreover, the
program also estimates the influence of single atoms on the overall activity of the molecule. Thus,
PASS can be used to systematically optimize a chemical structure with regard to desired effects, and
to minimize undesired side-effects. Along with another supporting program, PharmaExpert,
systematic data mining can be done to select those compounds out of a broad range of substances
that are the most promising ones for a certain purpose. Finally, a user can also load own data about
structure-activity relations as new training sets to create a proprietary structure-activity relationship
base for the program.
Multilevel Neighborhoods of Atoms (MNA) descriptors have been selected to describe the 2D
structural formulae of organic compounds. The molecular structure is represented in PASS by the set
of MNA descriptors of the 1st and 2nd levels. The prediction algorithm is based on Bayesian estimates
of probabilities for a compound to belong to the classes of active or inactive compounds,
respectively.
The current version of PASS (11.4.12) predicts 4,444 kinds of biological activity with an average
prediction accuracy of 95%.
The geneXplain platform: from functional genome analysis to biological process simulation
The geneXplain platform is an online toolbox and workflow management system for a broad range of
bioinformatic and systems biological applications. The individual modules are unified under a
standardized interface with a consistent look-and-feel and can be flexibly put together to
comprehensive workflows. The functionalities provided by the platform range from storage and
statistical processing of (e.g.) genomics or transcriptomics raw data through gene set enrichment
analysis, cluster analysis and identification of master regulators for a biological process to the
simulation of the dynamic behavior of a network.
A number of complex standard applications have been composed to pre-defined workflows. They can
be customized through a workflow management that is handled through a simple drag-and-drop
system also supporting the design of own workflows. Also, own scripts can be included into the
system and used in combination with pre-existing analyses.
It should be noticed that the platform provides a number of state-of-the-art modules; some of them
can be obtained free of charge, while others require licensing for small fee in order to guarantee
active maintenance and dynamic adaptation to the rapidly developing know-how in this field. We
explicitly encourage the community to contribute their own modules, which will be made available
and supported by the geneXplain platform.

2. One out of 24 million: How to computationally identify a new candidate drug against breast cancer
Apoptosis of breast cancer cells can be induced by the known drug RITA. However, at doses where it
efficiently induces p53-mediated apoptosis, it does not fully support p53-dependent inhibition of
survival genes, allowing some tumor cells to escape from apoptosis. In an attempt to find a way to
suppress survival genes as well, transcriptomics data of breast cancer cells were investigated using
the geneXplain platform technology. Comprehensive promoter and network analysis revealed a
number of potential master regulator genes. In a next step, the PASS approach was applied to
identify multitargeted compounds, i.e. substances that could inhibit with several master regulator
molecules at the same time. This way, and going subsequently through several filtering steps, 24
million chemical substances from the ChemNavigator library were virtually screened, leaving 64
compounds as potentially active ones. Of them, 26 turned out to be feasible for experimental testing,
out of which 2 substances exhibited highly specific effects.
Altogether, we conclude that our general strategy of integrated bioinformatic, systems biological and
cheminformatics analysis is highly suitable to identify promising lead structures for drug
development.
GeneXplain GmbH, Wolfenbuettel, Germany
GeneXplain is still a relatively young company, with the mission to provide a comprehensive platform
for bioinformatic, cheminformatic and systems biological tools. The raison d'être of this platform is to
assist translational research in the life sciences, mainly in the context of personalized medicine and
pharmacogenomics. We also make our expertise available to academic and commercial partners in
collaborative research projects.
Our strong commitment to science and research prompts us to seek for a new way of collaboration.
We feel that only by combining academic creativity with private commercial efforts, it will be
possible to generate a research platform that is both at the most recent state of the art as well as
reliable supported.
At present, geneXplain’s product portfolio comprises the cheminformatics tools PASS, PharmaExpert
and GUSAR, as well as the bioinformatics / systems biological geneXplain platform. In addition, we
distribute the software packages IMC (In-silico Molecular Cloning) and Genome Traveler outside
Japan for our partner in silico biology, inc in Yokohama.