• Environmental fate of Persistent Organic Pollutants (POPs)

POPs have been studied for their intrinsic tendency to global persistence and mobility (Long Range Transport, LRT) by applying multivariate methods (Principal Component Analysis and Multicriteria Descision making) and QSAR modelling, both by classification and regression models. A Global Half-life Index (GHLI) useful for the preventive screening and ranking of possible POPs, even before the synthesis, has been recently defined (ES&T, 2007).

P.Gramatica et al., Chemosphere, 2001, 43, 655; SAR and QSAR in Env. Res., 2002, 13, 205; Fresenius Environ. Bull. 2004, 13, 1204. P.Gramatica and E.Papa, Environ. Sci. Technol., 2007, 41, 2833.E.Papa and P.Gramatica, J.Mol. Graph. Model. 2008, 27, 59.

  • Screening and prioritization of Persistent, Bioaccumulative and Toxic compounds (PBTs) for regulatory purposes.

New tools, based on molecular structure and multivariate explorative methods, heve been developed for the ranking, identification and prioritization of PBTs. A new PBT index has been proposed.

E.Papa, and P.Gramatica Green Chemistry, 2010, 12, 836-843; [Selected as HOT Article]

  • Tropospheric reactivity with oxidants of Volatile Organic Compounds (VOCs)

The kinetic rate constants for the degradation reactions by tropospheric oxidants (hydroxyl and nitrate radicals , ozone) of heterogeneous Volatile Organic Compounds (VOCs) have been modelled by regression models (Multiple Linear Regression, MLR) using different kind of theoretical molecular descriptors, selected by Genetic Algorithms. Ranking of VOCs for their combined atmospheric degradability has been modelled by QSAR, in order to propose useful tools for the highlighting (also before the synthesis) of chemicals potentially persistent in troposphere.

P.Gramatica et al., Chemosphere, 1999, 38, 1371; SAR QSAR Env Res., 2002, 13, 743; Atm. Environ., 2003, 37, 3115; QSAR &Comb. Sci., 2003, 22, 364; J. Chem. Inf. Comp. Sci.. 2004, 44, 1794; Atm. Environ. 2004, 38, 6167; E.Papa and P.Gramatica, SAR QSAR Environ Res. 2008, 19, 655. P.P. Roy et al. J.Comput.Chem, 2011 32, 2386..

  • * Environmental partitioning of organic pollutants (pesticides and other emerging).
    * QSAR modeling of phys-chem properties: Koc, WSol, VP, H, etc.

The partitioning of pesticides into different environmental compartments and the dangerous leaching in ground-water depends mainly on the physico-chemical properties. The combination of two multivariate approaches (Principal Component Analysis and Hierarchical Cluster Analysis) and CART (Classification and Regression Tree) model was used to rank and classify pesticides of various chemical categories according to their distribution tendency in various media. A Leaching Index (LIN) and a Volatility Index (VIN) have been also proposed and modeled by MLR. The soil sorption coefficient, expressed as the ratio between chemical concentration in soil and in water, normalized to organic carbon (Koc), an important parameter when studying soil mobility and environmental distribution of chemicals, has been also modeled by externally validated highly predictive regression models, based on few simple molecular descriptors.[JMGM, 2007:Most cited paper in Computer Science]
In addition to Koc modelling, other common physico-chemical properties have been successfully modeled by combining different kinds of theoretical molecular descriptors: the basic physico-chemical properties of organic solvents, PCBs, esters, flame retardants, perfluorinated compounds, triazoles. The environmental distribution based on phys chem. properties and multivariate analysis by PCA is studied also for PFCs and (benzo) triazoles in the EU-FP7 Project CADASTER. A general classification of 152 organic solvents has been also proposed by applying k-nearest neighbor method and counterpropagation artificial neural network (CP-ANN), in particular Kohonen Maps.

P.Gramatica et al. Chemom. Intell. Lab. Syst., 1998, 40, 53-63; Trends Anal. Chem., 1999, 18, 461-471; Chemosphere, 2000, 41, 763; Chemosphere , 2002, 47, 947; Int. J. Environ. Anal. Chem. 2004, 84, 65; Fresenius Environ. Bull. 2004, 13, 1258. J. Mol. Graph. Model., 2007, 25, 755. [Most cited paper in Computer Sci]
E.Papa et al.QSAR& Comb. Sci., 2009, 28, 790; Molecular Informatics, 2011, 30, 232; B. Bhhatarai and P.Gramatica, Environ. Sci. Technol., 2011, 45, 8120; Water Research, 2011, 45,1463; B. Bhhatarai et al. Molecular Informatics, 2011, 30, 189.

  • Validation of QSAR models

A great attention is always devoted to the validation of the QSAR models developed by the group. Paola Gramatica, who is member of the OECD Principles (since the Setubal principles, 2002) has been involved in validation works for EU-JRC and OECD and has published papers and reports on validation (external, applicability domain, etc.) .). Recently a new proposal to use the Concordance Correlation Coefficient (CCC) of Lin for external validation of QSAR models has been published in JCIM.

A.Tropsha, P.Gramatica, V. K. Gombar QSAR &Comb. Sci., 2003, 22, 69-77. L. Eriksson et al. Environ. Health Perspectives 2003, 111, 1361. T. I. Netzeva et al. ATLA 2005, 33, 155. Vracko M. et al. SAR QSAR Environ. Res., 2006, 17, 265. P. Gramatica QSAR Comb.Sci. 2007, 26, 694[Most downloaded in 2007; 213 citations into March 2012]. Chirico and Gramatica JCIM, 2011, 51, 2320.
Reports for EU and OECD: P. Gramatica, 2004 under QSARs/documents/public access
Book chapters (see below)

  • Bioconcentration of organic pollutants in aquatic organisms

The Bioconcentration Factor (BCF) is an important parameter in environmental assessment as it is an estimate of the tendency of a chemical to concentrate and, consequently, to accumulate in an organism. The most common QSAR method, and the oldest, for estimating chemical bioconcentration is the establishment of correlations between BCF and chemical hydrophobicity using Kow, i.e. the n-octanol/water partition coefficient, but LogKow is an highly variable descriptors. Different models on BCF using theoretical molecular descriptors have been developed by the author group, with particular attention, as usual, to the external predictivity and the chemical applicability domain, according to the OECD principles.

P.Gramatica and E.Papa, QSAR &Comb. Sci., 2003, 22, 374; QSAR &Comb. Sci. 2005, 24, 953. E. Papa et al. Chemosphere, 2007, 67, 351.

  • Toxicity of organic pollutants

An innovative strategy for the selection of compounds with a similar toxicological mode of action was proposed in the study of chemical mixtures (PREDICT EU-Research Project). A representation of chemical structures for phenylureas and triazines by different molecular descriptors allowed a preliminary exploration of similarity based on Principal Components Analysis (PCA), MultiDimensional Scaling (MDS) and Hierarchical Cluster. The use of Genetic Algorithm to select the most relevant molecular descriptors in modeling toxicity data makes it possible to both develop good predictive QSAR toxicity models, and select the most similar phenylureas and triazines.

P.Gramatica et al. Chemosphere, 2001, 42, 873.

A proposal of new water quality objectives (WQO) for EEC priority List 1 (76/464/EEC) was realized by QSAR models and chemometric analysis of toxicity data.

M.Vighi, P.Gramatica, et al. Ecotox and Environ Safety, 2001, 49, 206.

The Duluth data set of toxicity data on Pimephales promelas was recently studied by the author group: new statistically validated MLR models were developed to predict the aquatic toxicity of chemicals classified according to their Mode Of Action (MOA). Also, a unique general model for Direct Toxicity Prediction (DTP model) was developed to propose a predictive tool with a wide applicability domain, applicable independently of a priori knowledge of the MOA of chemicals. Similarly, new QSAR models for esters' aquatic toxicity and a ranking for cumulative end-points were proposed.

E.Papa et al., J. Chem. Inf. Model 2005, 45,1256; Chemosphere 2005, Vol 58, 559.


In the EU-FP7 Project CADASTER toxicity of perfluorinated chemicals (PFCs) is studied: oral and inhalation rat and mouse toxicity have been modeled with the principal aim to select priority PFCs for experimental tests.

B. Bhatarai and P. Gramatica, Chem. Res. Toxicol., 2010, 23, 528; Molecular Diversity, 2011, 15, 467.


Similarly, oral acute toxicity and cytotoxic activity of fragrance materials in rodents for fragrances have been modeled:

E. Papa et al. SAR QSAR Environ Res., 2009, 20, 767.

  • Endocrine Disruptor Chemicals (EDCs)

Because of the large number of endocrine-disrupting chemicals (EDCs) in the environment, there is a great need for an effective tool of rapidly assessing ED activity in the toxicology assessment process and in the context of the new European REACH policy. Classification and regression QSAR models were developed to predict the estrogen receptor binding affinity based on a large data set of heterogeneous chemicals and theoretical molecular descriptors from DRAGON. The built OLS regression model, was validated comprehensively (internal and external validation, Y-randomization test) and all the validations indicate that the proposed QSAR model is robust and satisfactory. For the classification models, three nonlinear classification methodologies: Least Square Support Vector Machine (LS-SVM), Counter Propagation Artificial Neural Network (CP-ANN), and k-nearest Neighbor (kNN) were applied, by using four molecular structural descriptors as inputs. The models were also applied to about 58 000 discrete organic chemicals; about 77% were predicted not to bind to an estrogen receptor, but 8% should be prioritized for testing.
QSAR Studies on Selective Ligands for the Thyroid Hormone Receptor beta has been also performed.

H.Liu et al. Chem. Res. Toxicol., 2006, 19, 1540; J.Mol. Graph. Model. 2007, 26, 135; Bioorg. Med. Chem., 2007, 15, 5251; Chemosphere, 2008, 70, 1889; Comb. Chem. & H. T S. (special issue on "Machine learning for virtual screening") 2009, 12, 490.


Androgen receptor (AR) binders and pleiotropic endocrine disruptors (with double activity on AR and ER) have been also studied and modeled by regression and classification QSAR models for screening purposes.

J. Li and P. Gramatica, Molecular Diversity, 2010, 14, 687; J. Chem. Inf. Mod, 2010, 50, 861; SAR &QSAR in Environ Res, 2010, 21, 657.


In the EU-FP7 Project CADASTER the endocrine disruptor activity of flame retardants and PFCs is also studied.

E.Papa et al. Chem Res. Toxicol, 2010, 23, 946. S. Kovarich et al. J.Hazard. Mat., 2011, 190, 106. SAR&QSAR in Environ Res, 2012, DOI:10.1080/1062936X.2012.657235

  • Mutagenicity

The potential for mutagenicity of chemicals of environmental concern, like aromatic amines and PAHs, is of high relevance. With regard to this important topic, our group has published interesting MLR models, always verified for their external predictivity on new chemicals, for the Ames test results on amines and Nitro-PAHs. Externally validated classification models, by k-NN and CART, were also developed for the mutagenicity of benzo-cyclopentaphenanthrenes and ­ chrysenes, determined by the Ames test, and PAH mutagenicity, determined on human B-lymphoblastoid.

P.Gramatica et al., SAR QSAR Env Res., 2003, 14, 237; Ecotox. Environ. Safety, 2007, 66, 353; SAR QSAR Env Res.,2007, 18, 169;E. Papa et al. SAR QSAR Environ. Res., 2008 , 19,115.

  • Book chapters on Chemometric approach to QSAR modeling and validation

P. Gramatica Modelling Chemicals in the Environment, Chapter 17 in Drug Design Synthesis: quantitative approaches, D. Livingstone and A. M.Davis Eds., Royal Society Chemistry (RSC), Cambridge, U.K. 2012, 458-478.

P. Gramatica On the development and validation of QSAR models, in Computational Toxicology, B. Reisfeld, A. N. Mayeno Eds., Humana Press, Springer Science, New York, USA, 2011 in press.

P. Gramatica Chemometric Methods and Theoretical Molecular Descriptors in Predictive QSAR

Modeling of the Environmental Behaviour of Organic Pollutants, Chapter 12 in Recent Advances in QSAR Studies , T. Puzyn - J. Leszczynski - M.T.D. Cronin Eds., (Challenges and Advances in Computational Chemistry and Physics), Springer-Verlag New York Inc, 2009.

R. Todeschini, V. Consonni, and P. Gramatica Chemometrics in QSAR. In: Brown S, Tauler R, Walczak R (eds.) Comprehensive Chemometrics, volume 4, pp. 129-172 Oxford: Elsevier. 2009

Studies on the information related to particular descriptors (WHIM of shape) have been performed.

P.Gramatica, QSAR & Comb. Sci., 2006, 25, 327.


  • Application of Chemometric Methods to environmental problems

The group has also collaborated with other research groups for the interpretation of environmental data (metal accumulation by mosses as bioindicators, environmental sustainability parameters, leaching of pesticides in Uzbekistan rivers, identification of structurally representative and biodegradable antibiotics) by applying different chemometric explorative methods of multivariate analysis.

P.Gramatica et al. Environ. Sci. Pollut. Res, 2006, 13, 28 ; Fresenius Environ. Bull., 2006, 15, 731. E.Papa et al. Water Res. 2004, 38 , 3485 ; Environ. Sci. Technol., 2007, 41, 1653. S. Bastianoni , et al. J. Environ. Manag., 2008, 86, 419

  • EU-Projects on Mixture toxicity (PREDICT and BEAM) and Atmospheric pollutant toxicity (UNARO)

P.Gramatica has participated to EU-Projects on mixture toxicity (PREDICT and BEAM), as scientific responsible of Insubria Unit, and collaborated with Milano-Bicocca University in the UNARO project on atmospheric pollution. In the three projects, QSAR models and chemometric tools have been developed and applied.

Mixture Toxicity: P. Gramatica et al. Chemosphere, 2001, 42, 873; M. Vighi, P. Gramatica, et al. Ecotox and Environ Safety, 2001, 49, 206. M.Faust et al. Aquatic Toxicology, 2001, 56, 13; Aquatic Toxicology 2003, 63, 43. H. Walter et al. Ecotoxicology, 2002, 11, 299; M.Vighi, et al. Ecotox. Environ. Safety, 54, 2003, 139-150. Backhaus, T. et al. Continental Shelf Research 2003, 23, 1757; Environ. Toxicol. Chem. 2004, 23, 258. Atmospheric Pollution: P. Gramatica et al. Fresenius Environ. Bull. 2002,11, 757; Analysis in air, toxicology and QSAR modeling with nitrophenols. In: Air Pollution X, WIT Press, A. Brebbia, J. F. Martin-Duque eds, Southampton, U.K., pp 731-740. 2002.

  • Miscellaneous and collaborations for Chemical Design

Different works, which are not classifiable in the above topics, are also performed. The majority of them are collaborations with other research groups for chemical design.

P. Gramatica, Chemistry Today 2001, 18-24. T.Benincori, et al. Chem. Materials, 2001, 13, 1665. E. Bolzacchini, Bioconjugate Chemistry, 1999, 10, 332. P.Gramatica et al. Annali di Chimica, 2005, 95, 199. M. Pavan et al. In "Partial Order in Chemistry and Environmental Sciences" (R. Brüggemann and L. Carlsen Ed.), Chap. 3, 181, Springer-Verlag, 2006. L. Nizzetto et al. Environ. Sci. Technol., 2006, 40, 6580. S. Banfi et al. J. Med. Chem. 2006, 49, 3293. E. Papa et al. Environ. Sci. Technol., 2007, 41, 1653. J. H. Kim, P. Gramatica et al. SAR QSAR Environ. Res., 2007, 18, 729. J. Li et al. J. Comput. Chem., 2008, 29, 2636 ; J. Comput. Chem. 2010, 31, 973. H. Zhu et al. J.Chem.Inf.Model, 2008, 48, 766. I. Sushko et al., J. Chem. Inf. Model., 2010, 50, 2094. B. Lei et al., Atmos. Environ, 2010, 44, 2954. L.Xi et al. Chem. Engin. J., 2010, 163, 195. B. Bhhatarai et al. Molecular Informatics, 2010, 29, 511. P. Gramatica et al. J. Biol. Inorg Chem. 2010, 15, 1157