../index.html

Lecture of Prof. J. Gasteiger


Self-Organizing Neural Networks in Drug Design

Neural Networks

Supervised Learning

Unsupervised Learning

Somatosensory
Cortex

Somatosensory
Cortex

Architecture of the Kohonen Network

Simulation of Infrared Spectra

Operating a Counterpropagation (CPG) Network

Kohonen Networks in Drug Design

A Chemical Reaction

Representation of Structure

Dopamine and Benzodiazepine Agonists

Topological Autocorrelation Vector

PETRA

Topological Autocorrelation Vector

Classification of Biologically Active Compounds

Classification of Biologically Active Compounds

Structures Assigned to Neuron (3,9)

Dataset: The Hydantoin Library

Autocorrelation of Molecular Surface Properties

Autocorrelation of Molecular Surface Properties

Clustering the Hydantoin Library

Clustering the Hydantoin Library

Classification Rule

Best Classification Model

Discriminating Toxicants by Mode of Action with Neural Networks

Toxicant classes – Modes of Action

Questions

Description of Molecules

Classification According to Mode of Action
(1 output layer)

Classification According to Mode of Action
(1 output layer, centroids)

Classification According to Mode of Action
(9 output layers)

Estrogenic Compounds

Inhibitors of Acetylcholinesterase

Thiol-Alkylating Agents

Polar Non-specific Toxicants

Non-polar Non-specific Toxicants

Results

Hierarchy in Structure Representation

Structure Representation

Visualization of Kohonen Mapping
(KMAP 4.0)

Acknowledgements

References