Molecular Property Descriptors Research Articles

Ameliorative effects of Artemisia and Echinacea extracts against hepato and cardiotoxicity induced by DMBA on albino rats: experimental and molecular docking analyses

BackgroundHerbal therapy for healing disease has many advantages than drugs. This study investigates the protective efficacy of Artemisia annua (Art) and Echinacea pupurea (Ech) extracts against 7, 12-dimethylbenz (α) anthracene (DMBA) toxicity.ResultsDMBA-treated rats showed a significant increase in the level of serum ALT, AST, LDH and CKMB, also reduction in body weight gain (BWG) ℅, HB, WBCs, RBCs and platelet counts, in addition to histopathological and ultrastructural alterations. Rats treated with Art or Ech after DMBA showed little improvements in the biochemical, hematological, histopathological, ultrastructural and molecular docking results than before DMBA.ConclusionsThis study suggested the ameliorative effect of Ech and Art due to their antioxidant properties, but Ech and Art were more effective if they are given before than after DMBA administration and the marked effect against DMBA toxicity with Ech before DMBA exposure. Also, the molecular docking, molecular properties descriptors, and pharmacoinformatic studies of constituents of extract from Artemisia annua L. and Echinacea purpurea L. exhibited that all studied compounds have better ADMET and physicochemical properties, especially compounds extract from Echinacea purpurea L.Graphical

Partitioning Pattern of Natural Products Based on Molecular Properties Descriptors Representing Drug-Likeness

A cheminformatics procedure for a partitioning model based on 135 natural compounds including Flavonoids, Saponins, Alkaloids, Terpenes and Triterpenes with drug-like features based on a descriptors pool was developed. The knowledge about the applicability of natural products as a unique source for the development of new candidates towards deadly infectious disease is a contemporary challenge for drug discovery. We propose a partitioning scheme for unveiling drug-likeness candidates with properties that are important for a prompt and efficient drug discovery process. In the present study, the vantage point is about the matching of descriptors to build the partitioning model applied to natural compounds with diversity in structures and complexity of action towards the severe diseases, as the actual SARS-CoV-2 virus. In the times of the de novo design techniques, such tools based on a chemometric and symmetrical effect by the implied descriptors represent another noticeable sign for the power and level of the descriptors applicability in drug discovery in establishing activity and target prediction pipeline for unknown drugs properties.

Artificial neural network approach for predicting blood brain barrier permeability based on a group contribution method

Background and ObjectiveThe purpose of this study was to develop a quantitative structure-activity relationship (QSAR) model for the prediction of blood brain barrier (BBB) permeability by using artificial neural networks (ANN) in combination with molecular structure and property descriptors. MethodsUsing a database composed of 300 compounds, 52 structure descriptors obtained based on the universal quasichemical functional group activity coefficients (UNIFAC) group contribution method and the selected 8 molecular property descriptors were used as the network inputs, whereas logBB values of compounds constituted its output. ResultsThe correlation coefficient R of the constructed prediction model, the relative error (RE) and the root mean square error (RMSE) was 0.956, 0.857, and 0.171, respectively. These indicators reflected the feasibility, robustness and accuracy of the prediction model. Compared with the previously published results, a significant improvement in the predictions of the proposed ANN model was observed. ConclusionsANN model based on the group contribution method could achieve a satisfactory performance for logBB prediction.

Cytotoxic activity, molecular docking, pharmacokinetic properties and quantum mechanics calculations of the brown macroalga Cystoseira trinodis compounds

In this study, nine compounds were isolated, eight of them were isolated for the first time from Cystoseira trinodis. The biological activity of the extract, fractions and pure compounds was evaluated. The antimicrobial activity was investigated against 3 fungi species, 3 gram + ve and 3 gram –ve bacteria. The crude extract and fractions showed moderate inhibition against some of the tested microorganisms, especially the butanol fraction exhibited the maximum inhibition zone against Salmonella typhimurium (16 ± 0.60 mm). Cytotoxicity was evaluated against HepG-2 and MCF-7 cell lines. Hexane fraction exhibited the highest cytotoxic effect against HepG-2 and MCF-7 cell lines with an IC50 value of 14.3 ± 0.8 and 19.2 ± 0.7 µg/ml, respectively with compared to other fractions. The isolates were identified as octacosanoic acid (1), glyceryl trilinoleate (2), oleic acid (3), and the epimeric mixture of saringosterols (4, 5), β-sitosterol (6), glycoglycerolipid (7) and a mixture of kjellmanianone and loliolide (8, 9) by spectroscopic analysis. Among the all tested compounds kjellmanianone and loliolide mixture exhibited significant cytotoxic activity with an IC50 value of 7.27 µg/ml against HepG-2 cells. The major and minor constituents of the extract and fractions were identified using GC-MS analysis. Molecular docking analysis confirmed that most of the studied compounds especially compounds 8 and 9 strongly interact with TPK and VEGFR-2 with highest binding energies supported that the high cytotoxicity of these compounds against human hepatocellular cancer in the experimental part. The energetic, geometric and topological properties of compounds 8 and 9 binding with cytosine base were computed by DFT methods. Molecular properties descriptors, bioactivity score and ADMET analysis confirmed that most of the studied compounds especially compounds 8 and 9 exhibit significant biological activities and have a better chance to be developed as drug leads. Communicated by Ramaswamy H. Sarma

QSAR and molecular docking studies of lethal factor protease inhibitors against Bacillus anthracis

ABSTRACTBacillus anthracis is considered as a biological warfare agent because it is the causative agent of the serious infectious anthrax disease. Delay in treatment leads to lethal factor-mediated toxaemia which is very critical due to lack of therapeutic options. Consequently, attempts have been made to discover potent lethal factor (LF) protease inhibitors such as small-molecule synthetic 2-thio-1,3-thiazolidine-4-one (rhodanine) compounds. But computed descriptor-based quantitative structure–activity relationship (QSAR) and drug design studies on such aspect are poorly represented. Therefore, an attempt was made for developing QSAR models using structural descriptors for 1,3-thiazolidine-4-one compounds. The models were developed on a series of 49 LF protease inhibitors using the combination of constitutional, functional group, atom-centred fragment and molecular property descriptors. The best QSAR model included four variables, namely, C-040, nR05, GVWAI-80 and ALOGP that correlated well with the anti-LF protease activity with a good correlation coefficient (r = 0.870) of good statistical significance (F4, 29 = 14.09 (α = 0.001) F4, 29 = 6.19). This model was also validated and explained 58.1% of variances of the Bacillus anthracis inhibitory activities of the studied compounds with r2pred = 0.710 which denotes external predictability. Finally, molecular docking was carried out to predict the mode of binding of some highly active congeneric compounds. It was shown that VAL 1403 is an important residue for phenyl ring. TYR 1456 and HIS 1418 are responsible for interaction with the rhodanine nucleus. Therefore, these residues are considered responsible for the inhibition of LF protease anthrax and can predict significant dimension of essential structural features of these inhibitors to evaluate, screen and help priorities of the synthesis of the candidates against anthrax bioterrorism.

ToxicBlend: Virtual screening of toxic compounds with ensemble predictors

Timely assessment of compound toxicity is one of the biggest challenges facing the pharmaceutical industry today. A significant proportion of compounds identified as potential leads are ultimately discarded due to the toxicity they induce. In this paper, we propose a novel machine learning approach for the prediction of molecular activity on ToxCast targets. We combine extreme gradient boosting with fully-connected and graph-convolutional neural network architectures trained on QSAR physical molecular property descriptors, PubChem molecular fingerprints, and SMILES sequences. Our ensemble predictor leverages the strengths of each individual technique, significantly outperforming existing state-of-the art models on the ToxCast and Tox21 toxicity related bioactivity-prediction datasets. We provide free access to molecule bioactivity prediction using our model at http://toxicblend.owkin.com.

Baird’s Rule in Substituted Fulvene Derivatives: An Information-Theoretic Study on Triplet-State Aromaticity and Antiaromaticity

Originated from the cyclic delocalization of electrons resulting in extra stability and instability, aromaticity and antiaromaticity are important chemical concepts whose appreciation and quantification are still much of recent interest in the literature. Employing information-theoretic quantities can provide us with more insights and better understanding about them, as we have previously demonstrated. In this work, we examine the triplet-state aromaticity and antiaromaticity, which are governed by Baird’s 4n rule, instead of Hückel’s 4n + 2 rule for the singlet state. To this end, we have made use of 4 different aromaticity indexes and 8 information-theoretic quantities, examined a total of 22 substituted fulvene derivatives, and compared the results both in singlet and triplet states. It is found that cross-correlations of these two categories of molecular property descriptors enable us to better understand the nature and propensity of aromaticity and antiaromaticity for the triplet state. Our results have not only demonstrated the existence and validity of Baird’s rule but also shown that Hückel’s rule and Baird’s rule indeed share the same theoretical foundation because with these cross-correlation patterns we are able to distinguish them from each other simultaneously in both singlet and triplet states. Our results should provide new insights into the nature of aromaticity and antiaromaticity in the triplet state and pave the road toward new ways to quantify this pair of important chemical concepts.

Introducing a computational method to estimate and prioritize systemic body exposure of organic chemicals in humans using their physicochemical properties

This report describes a computational method developed to predict systemic exposure (s-exposure), chemical disposition {(CD) intestinal absorption, transport, membrane permeability, distribution, sequestration, phospholipidosis and toxicokinetics} of organic chemicals in humans. The method qualitatively and quantitatively estimates a chemical's CD activity profile based upon computed molecular descriptor properties (descriptors), and it facilitates in silico signal-detection of data-gaps, prioritization, risk-ranking, read-across, and re-assessments (if mandated) of large sets of chemicals in a safety evaluation setting. The investigation used a reference set of 2372 marketed human pharmaceuticals to define decision rules for an optimal chemical space (OCS) in which chemicals have high s-exposure, good CD, and a potential for chemical toxicity (CT); conversely, chemicals outside the OCS have low s-exposure, poor CD into the body, and low potential for CT. The method requires computation of 29 descriptors, identification of OCS molecular descriptor property violations (descriptor_violations), and alignment of descriptor_violations with specific decision rules for individual CD endpoint activities. The investigation predicted the CD activities of food and cosmetic preservatives, ingredients in GRAS (generally recognized as safe). Notices submitted to the FDA, reference pharmaceuticals, and it provides prioritization metrics and indices that facilitate prioritization of chemical in silico computed CD activities.

Theoretical Studies of Intracellular Concentration of Micro-organisms\u2019 Metabolites

With the rapid growth of micro-organism metabolic networks, acquiring the intracellular concentration of microorganisms’ metabolites accurately in large-batch is critical to the development of metabolic engineering and synthetic biology. Complementary to the experimental methods, computational methods were used as effective assessing tools for the studies of intracellular concentrations of metabolites. In this study, the dataset of 130 metabolites from E. coli and S. cerevisiae with available experimental concentrations were utilized to develop a SVM model of the negative logarithm of the concentration (-logC). In this statistic model, in addition to common descriptors of molecular properties, two special types of descriptors including metabolic network topologic descriptors and metabolic pathway descriptors were included. All 1997 descriptors were finally reduced into 14 by variable selections including genetic algorithm (GA). The model was evaluated through internal validations by 10-fold and leave-one-out (LOO) cross-validation, as well as external validations by predicting -logC values of the test set. The developed SVM model is robust and has a strong predictive potential (n = 91, m = 14, R2 = 0.744, RMSE = 0.730, Q2 = 0.57; R2p = 0.59, RMSEp = 0.702, Q2p = 0.58). An effective tool could be provided by this analysis for the large-batch prediction of the intracellular concentrations of the micro-organisms’ metabolites.

In silico receptor-based drug design of X,Y-benzenesulfonamide derivatives as selective COX-2 inhibitors

COX-2 is a widely studied biological target, since its activity is directly related to the inflammation response. The design of COX-2 selective inhibitors is an ongoing topic in drug design. We performed a quantitative structure–activity relationship and docking studies over a series of benzenesulfonamide derivatives on their inhibition towards COX-1 and COX-2, in order to rationalize their selectivity towards COX-2. Constitutional, topological and molecular property descriptors for the QSAR models and molecular docking calculations were employed. The mathematical model highlighted that lipophilic character and size are the most important features for COX-2 inhibition by benzenesulfonamides. A second QSAR model revealed that the dipole moment, the number of hydrogen bond donors and lipophilicity descriptors of benzenesulfonamides are crucial for their binding to COX-1. Moreover, artificial neural networks were employed to improve the prediction power of the COX-1 inhibition QSAR model. In this sense, we proposed new selective potential inhibitors by introducing different halogens into the benzenesulfonamide scaffold, improving their interactions with key residues of COX-2.

Quantitative Structure–Property Relationship Predictions of Critical Properties and Acentric Factors for Pure Compounds

Knowledge of critical constants and phase boundary pressure properties is essential to understanding thermodynamic behavior of substances and is often required in practical process design applications. Where critically evaluated data are unavailable, a quantitative structure–property relationship (QSPR) regression method can be used to relate molecular properties (descriptors) to properties of interest. The relationship is trained and tested using existing critically evaluated data and is dynamic; as new data become available, the relationship can be updated to reflect changes. In this work, we use support vector regression (SVR) to develop estimation methods for critical properties and acentric factors based on critically evaluated data for over 900 pure compounds. From three-dimensional geometry and connectivity information, we calculate over 500 descriptors for each compound. A matrix of descriptor values defines the input vectors for SVR, whereas critically evaluated data for critical temperature, the...

Predictors for gases of high electrical strength

We present an efficient method of predicting the electric strength (ES) of insulating gases in comparison to sulfur hexafluoride. Different molecular properties (descriptors) of a comprehensive set of 67 predominantly electronegative or electron attaching molecules were calculated ab-initio by means of density functional theory (DFT). For the same list of molecules, we compiled a data set of experimental values for the ES and descriptors. We analyzed the data by statistical methods and observed strong correlations between the ES and certain predictors, which are simple functions of selected DFT-calculated descriptors. In addition, we applied the same statistical method to the boiling point of the gas, and we observed strong correlations as well. We demonstrate our method by predicting the ES for a few unreported molecules.

Nearly no Scoring Function Without a Hansch-Analysis.

Corwin Hansch is well-known as the father, inventor, and promoter of quantitative structure-activity relationships. Usually, QSAR is seen as a ligand-based design method correlating molecular structure or property descriptors to biological activity. QSAR is seldom mentioned in relation to structure-based approaches, although it is the centerpiece of nearly every empirical scoring function. QSAR techniques are applied on various levels, from the fitting of scoring terms to biological affinity data in empirical scoring functions up to the fine-tuning of individual aspects of protein-ligand interactions. In the following, we report on current findings for our scoring approach HYDE, which are based upon both the idea of QSAR and Hansch's historical logP data. We relate the molecular surface area of 594 diverse compounds to their experimental octanol/water partition coefficients aiming at new insights in hydrogen bonding and dehydration energies of solutes. Donors and acceptors which are far from each other contribute with nearly constant increments to the logP value. The solubility in the aqueous phase is however not increased with the number of hydrogen bonds a polar group is able to form. Although signs are found that these facts have been known for many years, they have implications for modern scoring function design.

Prediction of Aqueous Solubility of Organic Solvents as a Function of Selected Molecular Properties

Prediction of Aqueous Solubility of Organic Solvents as a Function of Selected Molecular Properties The aqueous solubility of 66 organic solvents was examined. As an organic solvent molecule is not always simple (single-carbon) organic species, the aqueous solubility was tested as a function of some selected molecular properties (descriptors); namely and using VolSurf+ notation, CP, D1, D2, D3, FLEX, G, HSA, LogKow, MW, PHSAR, POL, PSA, PSAR, R, S, V, W1, W2, and W3. In general, the organic solvent solubility data can be best described by the two molecular descriptors LogKow and R, with PHSAR as the third refining or tuning-up factor (weight function).

Predictive Models for Maximum Recommended Therapeutic Dose of Antiretroviral Drugs

A novel method for predicting maximum recommended therapeutic dose (MRTD) is presented using quantitative structure property relationships (QSPRs) and artificial neural networks (ANNs). MRTD data of 31 structurally diverse Antiretroviral drugs (ARVs) were collected from FDA MRTD Database or package inserts. Molecular property descriptors of each compound, that is, molecular mass, aqueous solubility, lipophilicity, biotransformation half life, oxidation half life, and biodegradation probability were calculated from their SMILES codes. A training set (n = 23) was used to construct multiple linear regression and back propagation neural network models. The models were validated using an external test set (n = 8) which demonstrated that MRTD values may be predicted with reasonable accuracy. Model predictability was described by root mean squared errors (RMSEs), Kendall's correlation coefficients (tau), P-values, and Bland Altman plots for method comparisons. MRTD was predicted by a 6-3-1 neural network model (RMSE = 13.67, tau = 0.643, P = 0.035) more accurately than by the multiple linear regression (RMSE = 27.27, tau = 0.714, P = 0.019) model. Both models illustrated a moderate correlation between aqueous solubility of antiretroviral drugs and maximum therapeutic dose. MRTD prediction may assist in the design of safer, more effective treatments for HIV infection.

Spherical harmonics coefficients for ligand-based virtual screening of cyclooxygenase inhibitors.

BackgroundMolecular descriptors are essential for many applications in computational chemistry, such as ligand-based similarity searching. Spherical harmonics have previously been suggested as comprehensive descriptors of molecular structure and properties. We investigate a spherical harmonics descriptor for shape-based virtual screening.Methodology/Principal FindingsWe introduce and validate a partially rotation-invariant three-dimensional molecular shape descriptor based on the norm of spherical harmonics expansion coefficients. Using this molecular representation, we parameterize molecular surfaces, i.e., isosurfaces of spatial molecular property distributions. We validate the shape descriptor in a comprehensive retrospective virtual screening experiment. In a prospective study, we virtually screen a large compound library for cyclooxygenase inhibitors, using a self-organizing map as a pre-filter and the shape descriptor for candidate prioritization.Conclusions/Significance12 compounds were tested in vitro for direct enzyme inhibition and in a whole blood assay. Active compounds containing a triazole scaffold were identified as direct cyclooxygenase-1 inhibitors. This outcome corroborates the usefulness of spherical harmonics for representation of molecular shape in virtual screening of large compound collections. The combination of pharmacophore and shape-based filtering of screening candidates proved to be a straightforward approach to finding novel bioactive chemotypes with minimal experimental effort.

Chemogenomic databases: Construction, search and analysis

Over the past 10–15 years, the advent of high-throughput technologies developed by the pharmaceutical industry and related disciplines has produced large databases of drug efficacy, gene/protein expression levels, mutational status, as well as molecular structure information of potential drugs and drug targets which further our understanding of normal versus disease states and are useful for drug discovery. With the recent availability of large public databases, notably those from US government agencies such as the National Center for Biotechnology Information and the National Cancer Institute, the emerging field of chemogenomics is providing abundant opportunities for innovative new techniques for statistical analysis, data integration and applications of large-scale datamining. This is the first of two issues of Statistical Analysis and Data Mining to be published in 2009 focused on statistical methods for chemogenomics. The review by Liu and Verducci provides a good introduction to the field of chemogenomics as well as an overview of the many statistical, bio/chemo-informatic, and experimental subdisciplines used in modern drug discovery. Several papers in this issue describe new statistical methods for relating molecular structure to drug efficacy. Because compounds that have similar molecular structure often have similar biological properties, chemical similarity searching, on the basis of features extracted from the molecular structure, has long been used to prioritize compounds for testing. The article by Liu and Verducci introduces methods of chemical similarity searching that typically start with an active compound as a reference structure and aim to retrieve additional active compounds from a database. If instead of a single compound the search is based on multiple active compounds, the proportion of active compounds retrieved from the database can be significantly enriched. Turbo similarity searching (TSS) assumes that the nearest neighbors of the single reference compound provided by the user are active compounds and automatically uses them to conduct a multiple reference compound similarity search. Previous studies have shown that TSS can improve retrieval enrichment compared to a traditional similarity search. In this issue, Gardiner et al. compare the effectiveness of TSS over a variety of databases and use alternative structural descriptors for computing structural similarity. Scheiber et al. present a method for dealing with multiple activity values (i.e. from a series of assays) with the aim of explaining the assay differences from a chemical perspective. The key idea is to create meta-categories from the differences in assay values and use the new (ordinal) categories as dependent variables to identify chemical properties or molecular substructures associated with the difference in profiles. In the chemogenomics setting, this technique could be used to identify molecular features common to compounds for which activity levels are associated with gene/protein expression signatures. The effectiveness of a molecular descriptor set for identifying active compounds from a similarity search depends on the compound class. This dependency makes it difficult to select the most suitable descriptor set for a given search. Vogt et al. describe a method that combines a Bayesian scoring scheme for selection of structural and molecular property descriptors with an information-theoretic method for predicting recall rates. The procedure allows for the selection of search methods most likely to be successful for a given compound class and target database. Gardner-Lubbe et al. discuss the advantages of biplots for visualization and analysis of microarray data. Biplots are typically used in conjunction with principal component analysis. However, the authors show that PCA biplots do not provide optimal data separation when used for exploring the differences between treatments and differentially expressed genes. As an alternative, they suggest biplots based on analysis of distance and illustrate its effectiveness in separating gene expression samples from three treatment groups. As a whole, the papers in this issue take the reader from the fundamentals of how chemogenomic databases are constructed and searched, to a clever application of statistical analysis used to draw insight from these. The works reflect the best spirit of data mining for ideas, not just facts. The authors are thanked and congratulated.

Predicting the similarity search performance of fingerprints and their combination with molecular property descriptors using probabilistic and information theoretic modeling

AbstractSimilarity searching is currently one of the most widely applied approaches to computationally screen large databases for novel active compounds, and molecular fingerprints are among the most popular search tools. Fingerprint searching has recently also been applied in chemical biology to identify compounds that are selective for a target within a group of related ones. In general, fingerprints are bit string representations of molecular structure and properties but their design, size, and complexity often vary substantially. Like essentially all similarity search tools, fingerprints display a strong compound class dependence in their ability to identify active molecules and distinguish them from other database compounds. In practical applications, this limitation makes it very difficult to select or prioritize fingerprints that are most suitable for a given search problem. We have previously (i) devised a Bayesian‐scoring scheme to combine fingerprints and molecular property descriptors for similarity searching and (ii) developed an information‐theoretic approach to predict active compound recall rates for fingerprint searching. Herein, we combine these methods and present an approach for the prediction of compound recall in search calculations using Bayesian screening with molecular property descriptors, fingerprints and their combination. For practical similarity search applications, this analysis is highly relevant because it makes it possible to identify search methods that are most likely to be successful for a given compound activity class and screening database. Copyright © 2009 Wiley Periodicals, Inc. Statistical Analysis and Data Mining 2: 123‐134, 2009

In Silico Screening of Chemicals for Genetic Toxicity Using MDL-QSAR, Nonparametric Discriminant Analysis, E-State, Connectivity, and Molecular Property Descriptors

ABSTRACTGenetic toxicity testing is a critical parameter in the safety assessment of pharmaceuticals, food constituents, and environmental and industrial chemicals. Quantitative structure-activity relationship (QSAR) software offers a rapid, cost-effective means of prioritizing the genotoxic potential of chemicals. Our goal is to develop and validate a complete battery of complementary QSAR models for genetic toxicity. We previously reported the development of MDL-QSAR models for the prediction of mutations in Salmonella typhimurium and Escherichia coli (); this report describes the development of eight additional models for mutagenicity, clastogenicity, and DNA damage. The models were created using MDL-QSAR atom-type E-state, simple connectivity and molecular property descriptor categories, and nonparametric discriminant analysis. In 10% leave-group-out internal validation studies, the specificity of the models ranged from 63% for the mouse lymphoma (L5178Y-tk) model to 88% for chromosome aberrations in vivo. Sensitivity ranged from a high of 74% for the mouse lymphoma model to a low of 39% for the unscheduled DNA synthesis model. The receiver operator characteristic (ROC) was ≥2.00, a value indicative of good predictive performance. The predictive performance of MDL-QSAR models was also shown to compare favorably to the results of MultiCase MC4PC () genotoxicity models prepared with the same training data sets. MDL-QSAR software models exhibit good specificity, sensitivity, and coverage and they can provide rapid and cost-effective large-scale screening of compounds for genotoxic potential by the chemical and pharmaceutical industry and for regulatory decision support applications.

Some molecular descriptors for non-specific chromosomal genotoxicity based on hydrophobic interactions

A concept relating the lipophilicity of chemicals with their genotoxicity on a chromosomal level had been generated by Schultz and Onfelt (Chem Biol Interact 126:97-123, 2000). It was shown that aneuploidy in Chinese hamster V79 cells was elicited by lipophilic chemicals at concentrations related to their hydrophobicity (log P), whereas toxicants with a specific mode of action acted at concentrations consistently lower than predicted based on log P. We have now combined available data sets on aneuploidy/micronucleus formation with procedures used in QSAR modelling, in order to find new molecular descriptors for modelling non-specific chromosomal genotoxicity, and to optimise combinations thereof. Molecular structures of 26 chemicals, including steroids, were converted into single 3D models using Corina (version 3.20), and 11 descriptors of molecular properties were calculated. The data of 16 compounds assigned to a non-specific mode of action were imported into the QSAR module of the software package Cerius(2) (version 4.10). Applying genetic function approximation (GFA), linear equations were set up relating molecular descriptors with experimental concentrations at which doubling of micronuclei occurred in V79 cells (exp -log C). The number of variables (molecular descriptors) was limited to a maximum of three, and linear and quadratic terms were allowed. Based on the descriptions provided by the GFA procedure, log P was the most suitable single property to describe non-specific genotoxicity [r (2 ) = 0.88], confirming the original concept of Schultz and Onfelt. Using more descriptors (up to three in combination) resulted in an optimization of correlations up to r (2 )= 0.97. Such optimal correlation coefficients were obtained by combinations (a) of the numbers of hydrogen bond acceptors, the polar surface and total surface areas of molecules on one hand, and by (b) the dipole moment, polar surface and total surface descriptors on the other hand. In essence, the relation of polar surface to the total molecular surface appears pivotal to determine the non-specific chromosomal genotoxicity of lipophilic compounds.