Abstract: MATH/CHEM/COMP 2002, Dubrovnik, June 24-29, 2002



Prediction of anti-HIV-1 Activity for Porphyrin Derivatives using Artificial Neural Networks (ANN) and

Projection of Latent Structures (PLS) Techniques


Karoly Heberger, Rozalia Vanyur, and Judit Jakus


Institute of Chemistry, Chemical Research Center, Hungarian Academy of Sciences, P.O. Box 17, H-1525 Budapest, Hungary




Quantitative Structure-Activity Relationship (QSAR) is widely used in rational drug design. Earlier, we investigated the usage of 3D descriptors, variable selection and predictive models for tumoricidal activity and accumulation of photosensitizers of tetrapyrrol structure 1.

In this work, 21 compounds with similar structures (porphyrins) were examined as anti-HIV agents using comparative molecular field analysis (CoMFA) 2. Our aim was to develop predictive models for anti HIV activity using 3D-descriptors. The molecular structures were optimized using HyperChem 3, whereas the descriptors (84) were calculated using 3Dnet 4 program package.

Different feature selection and modeling methods (such as artificial neural networks, ANN and partial least squares, PLS) were compared considering their predictive performance.

The 3D QSAR models were carefully validated using

(i) shuffling the activity (response) values,

(ii) dividing the data (compounds) into training monitoring and prediction sets,

(iii) cross validation, and

(iv) external validation.

An ANN model with one hidden layer and three hidden nodes containing three descriptors (KMASS, DB, DBE) provided a Q2 = 0.77 (leave one out) and Q2 = 0.79 (external validation), whereas the best PLS-CoMFA with four PLS components had only a Q2 = 0.59 2.


1 R. Vanyur, K. Heberger, I. Kovesdi, J. Jakus, Prediction of tumoricidal activity and accumulation of photosensitizers in photodynamic therapy using multiple linear regression and artificial neural networks. Photochemistry and Photobiology 75 (2002) May.

2 A. K. Debnath, S. Jiang, N. Strick, K. Lin, P. Haberfield, A. R. Neurath, Three-dimensional structure-activity analysis of a series of porphyrin derivatives with anti-HIV-1 activity targeted to the V3 loop of the gp120 envelope glycoprotein of the human immunodeficiency virus type 1. J. Med. Chem. 37 (1994) 1099-1108.

3 HyperChem version 4.0, HyperCube Inc. Canada, 1997.

4 3DNET version 1.0, CompElit Ltd. Hungary, 1998.