Community Research and Development Information Service - CORDIS

Abstract

In the scheme of priority setting of dangerous chemicals, predictive methods like quantitative structure-activity relationships (QSAR) play an important role. A QSAR is a regression model that relates the variation in the biological activity (Y-matrix) within a homogenous series of compounds to their chemical and structural properties (X-matrix).

The present paper is assigned to investigate the usefulness of these X-variables in cases of QSARs and environmental modelling. The approach will be used in 2 examples, modelling both quantitative and qualitative biological responses. The first data set concerns classification of 512 compounds into 4 different classes: non-polar narcotics; polar narcotics; reactive chemicals; specifically acting chemicals. The classification was carried out by supervised clustering using partial least squares (PLS) discriminant analysis. In the second example, the octanol-water partition coefficient (LogP) was modelled for a set of 50 compounds. Both examples showed promising results.

Additional information

Authors: LINDGREN F, JRC Ispra (IT);NOUWEN J, JRC Ispra (IT);LOONEN H, JRC Ispra (IT);WORTH A, JRC Ispra (IT);HANSEN B, JRC Ispra (IT);KARCHER W, JRC Ispra (IT)
Bibliographic Reference: Paper presented: Assessing Low-Dose Effects of Air Pollutants, Pavia (IT), September 15-16, 1995
Availability: Available from (1) as Paper EN 39321 ORA
Record Number: 199512045 / Last updated on: 1996-01-15
Category: PUBLICATION
Original language: en
Available languages: en