Chemoinformatics and bioinformatics are relatively new terms, but they today have an innovative
and powerful influence on various important subjects such as drug discovery or elucidating of
important biological functions of enzymes. They also have a great deal to do with the current and
potential contributions of computer science to human health. Especially chemoinformatics has an
increasingly important role in drug discovery and drug design. In terms of this trend, the CBI team
has been developing chemical software for drug discovery and understanding the relationships
between chemical structure and biological activity.
In 2004, our team released onto the market a commercial software package called "PreADME",
intended to provide the most outstanding and practical guidance for drug discovery. This PreADME
had great meaning as the first drug discovery software package released for commercial purpose
in Korea, with an attempt to provide computational (or rational) solution to the early drug discovery.
However, the missions of our CBI in Research Institute of Bioinformatics & Molecular Design (BMD)
is like the following
1. Physico-chemical properties prediction of diverse chemicals
About 2,500 2D, 3D molecular descriptors, which represent various physico-chemical properties
of compounds such as boiling point, water solubility, etc., are calculated in our product, PreADME.
There are constitutional, topological, physico-chemical and geometrical descriptors that are useful
to the prediction of ADME/Tox. properties of drug candidates, for example the number of H- bond
acceptors, The number of rotatable bonds, calculated logP, topological polar surface area, molecular
connectivity indices, and so on. In addition, we are very interested in implementing and integrating
statistical computation, artificial intelligence like neural network and innovative computational
algorithm into our research purpose, especially drug discovery. As a result of our endeavors and
research, we have built robust and reliable predictive models of biological activities such as predicting
the ADME/Tox. properties of drug candidates.
2. In Silico High Throughput ADME/Tox. Screening
Recently, the emphasis in the drug discovery of many pharmaceutical companies has been
extended from finding bioactive molecules by the traditional QSAR to predicting the ADME/Tox.
(absorption, distribution, metabolism, excretion, and toxicity) of drug candidates in advance,
because a number of new drugs in the drug market have been failed due to poor pharmacokinetic
properties. The important goal of the CBI team is constructing predictive models of ADME/Tox.
and druglike properties of chemicals, finally intended for the rational drug design. Actually, we
developed powerful predictive models using artificial neural network (with Resilient
back-propagation (Rprop) method) in order to evaluate molecules as potential drug candidates,
especially by predicting their permeability in Caco-2 and MDCK cells for oral bioavailability and
by predicting their penetration through BBB (blood-brain barrier) and plasma protein binding for
drug distribution. Furthermore, we are interested in the prediction of physico-chemical properties
of chemicals that have a close connection to the bioavailability of therapeutic agents. In fact,
we had successfully made computer programs to calculate molecular descriptors that are closely
related to the ADME/Tox. properties. In summary, our ultimate goals are like the following:
Excretion prediction: Hepatic clearance, urinary and bile excretion
Pharmacokinetics prediction: Bioavailability (Blood level), volume of distribution, half life time,
dose response, microsomal statbility
Toxicity prediction: Mutagenicity (Ames test), carcinogenicity (rodent bioassay), Rat and
mouse oral LD50, Chronic LOAEL, hERG inhibition, skin sensitization, skin and eye irritation,
hepatoxicity
Drug-likeness prediction: Lipinski’s rule (Rule of Five), Drug-like rule based on CMC, MDDR, and
WDI, Leadlike rule
3. Chemical structure and property databases for drug design
We have constructed chemical databases including about 2,000,000 compounds and developed
various search engines for the drug discovery. Our chemical databases contains 2D/3D chemical
structures and their 2D/3D descriptors, for example, H-bond acceptor and donor, rotatable bond,
polar surface area, calculated logP, molecular connectivity, etc. for example. In our search engine,
not only simple search with chemical names, molecular weight and molecular formula but the
similarity search by 2D or 3D structures and 2D exact structure/substructure search are
prepared also. Conceptual design of organic chemical database Over 2 million 2D, 3D chemical structures
and their properties or information like common name Exact, substructure, similarity searching in chemical database 2D fingerprint representation of molecular structure and properties for quick chemical searching Design of viewer module (Bioinformatical support for drug-target discovery)
4. Chemical library design and analysis
Combinatorial chemistry and high-throughput screening (HTS) are now an established part of
the drug discovery process, but they are somewhat based on the probability of finding a hit in
screening experiments (using universal or general libraries spanning many chemical classes)
according to the number and variety of molecules screened. Because of the cost to synthesize
and screening an enormous number of chemicals, researchers are attempting to enhance the
information content of screening libraries, for example designing focused chemical libraries for
a particular purpose. We have developed powerful software to quickly calculate above 1,000
molecular descriptors of many molecules simultaneously, in order to design virtual combinatorial
libraries that includes enough molecular diversity. Also, implementing data minding technology
such as inferential statistics like PCA and artificial intelligence like genetic algorithm is of our
important interest. On the basis of such methods, we can integrate our software into the
development of lead drug candidates through chemical space navigation, or molecular diversity
analysis to optimally select molecule sets whose structures cover chemical space or span a
particular chemical property without the redundancy of the chemicals themselves. Our main goals are like the below: Molecular descriptors for Combi. Chem. and HTS library design: Fast topological descriptors :
Kier & Hall descriptors, information contents, E-state keys and diverse fingerprint keys,
BCUT values 2D, 3D fingerprint key design: 2D similarity, classification of atomic types, pharmacophore
definition Library selection rule: Considering molecular diversity, ADME/Tox, reagents cost, deconvolution Chemical similarity: Tanimodo coefficient and distance using BCUT and fingerprint descriptor
5. In silico prediction of environmental toxicity
Recently, many numbers of chemicals are synthesized and released to the environment without
sufficient consideration of their effect to the environment. Predicting chemical toxicity and fate in
the environment is one of most interest subject in many industrial, government, and NGO. Firstly,
we are developing a prediction model for acute fish toxicity based on our developed descriptors
and machine learning methods. And we expand our prediction model to vary endpoints in the
environmental toxicity. Our final goal is prediction of the environmental effects of chemicals at
different levels of biological organization and chemicals fate in the environment.
Chem/Bioinformatics (CBI)
Home>R&D> Chem/Bioinformatics (CBI) 
