Searching and Optimizing Lead Structures

A leadstructure is a chemical compound having a certain biological effect, e.g. blocking an enzyme. For drug discovery, such lead structures are modified to create more potent substances. The search for lead structures is a very basic and important task in drug design. One of the methods to do so is to screen substances from huge databases or combinatorially synthesized substances for their pharmaceutical value (so called High-Throughput-Screening, HTS). We aim to develop procedures that computationally support or simulate these screenings.

Assuming that similar compounds have similar properties, computer programs were developed during the preceding project ÄBAV that score the similarity of compounds contained in data bases. To do so, important molecular properties are coded ( descriptor ) and these descriptors are analyzed by e.g. Kohonen-Networks which are artificial neural networks that can automatically classify data. Another research activity are methods yielding quantitative results for a given biological effect, e.g. the concentration needed for a 50% inhibition of a given enzyme, for similar compounds ( Quantitative Structure-Activity Relationships, QSAR ).

We cooperate with the Computer Chemistry Center at the University of Erlangen-Nürnberg, the Institute for Pharmaceutical Chemistry at the University of Marburg, and the companies Merck, Solvay and Byk Gulden.

The University of Tübingen develops descriptors and a program system for the project.

Further information

  • The Java-based software package COFEA implements the concept of the Compressed Feature Matrix, enabling a fast and adaptive search for similarities, a pharmacophore development and the search for substructures
  • The Java- based 'computational chemistry' library JOELib is a data handling tool for converting molecules, SMARTS substructure search and other useful topics

Project partners

The SOL-project is supported by the German Federal Ministry of Education and Research (BMBF) , contact number 311681.


Building Neural Network Models for the Prediction of Aqueous Solubility and Partition Coefficient Based on 2D/3D descriptor sets optimized by a genetic algorithm
J. K. Wegner, A. Zell (2003)
Journal of Chemical Information and Computer Science (JCICS), accepted
The Compressed Feature Matrix - A fast method for feature based substructure search
S.F.Badreddin Abolmaali, Jörg K. Wegner, Andreas Zell (2003)
Journal of Molecular Modeling, accepted
Discovering new lead structures for Gd-based contrast agents using the Compressed Feature Matrix
Badreddin F. Abolmaali, Andreas Zell, T. J. Vogl, Nasreddin D. Abolmaali (2003)
European Congress of Radiology, accepted
The Compressed Feature Matrix - A novel descriptor for adaptive similarity search
S.F.Badreddin Abolmaali, Claude Ostermann, Andreas Zell (2003)
Journal of Molecular Modeling, published online
DOI 10.1007/s00894-002-0110-0 [HTML]
Compressed Feature Matrix - Ähnlichkeitsbewertung und Substruktursuche in großen Datenbanken
S.F.Badreddin Abolmaali, Andreas Zell (2001)
15. Molecular-Modelling Workshop, Darmstadt, 22.-23. Mai 2001
[Abstract] []
Einfluß der Gewichtung von Moleküldeskriptoren auf die Klassifizierung von Arzneistoffen
S.F.Badreddin Abolmaali, Andreas Zell (2000)
14. Molecular-Modelling Workshop, Darmstadt, 30.-31. Mai 2000
[Abstract] [Poster]
SOLVES - Ein verteiltes Programmsystem zur Suche und Optimierung von Leitstrukturen
Fred Rapp, Thomas Kleinöder, Andreas Dominik, Andreas Zell (2000)
14. Molecular-Modelling Workshop, Darmstadt, 30.-31. Mai 2000


Badreddin Abolmaali, Tel.: (07071) 29-78979, abolmaali at

Jörg Wegner, Tel.: (07071) 29-78970, wegnerj at

Fred-Reiner Rapp, Tel.: (07531) 84-4621, fred_rapp at