Quantitative structure-activity relationships (QSAR) for pesticide regulatory purposes / 1st ed.

Forewords p. xi
The DEMETRA Project: An Innovative Contribution to Regulatory QSAR p. xi
Reference p. xiv
Preface Emilio Benfenati and Mose Casalegno p. xv
The Pesticides and their Ecotoxicological Properties p. xv
Moving Forwards the use of QSAR to Predict Toxicological Properties p. xvi
The DEMETRA Project p. xvii
The Book Chapters p. xix
Acknowledgement p. xx
Disclaimer p. xx
References p. xxi
Chapter 1 QSARs for regulatory purposes: the case for pesticide authorization Emilio Benfenati and Mark Clook and Steven Fryday and Andy Hart p. 1
1 Overview of the Current Pesticide Authorization Procedure p. 1
1.1 Description of the current pesticide legislation (EU Directive 91/414/EEC) p. 1
1.2 Outline of the ecotoxicology tests required for pesticide authorization under 91/414/EEC p. 3
1.3 How frequently are certain studies submitted and how many studies are submitted to address an Annex point? p. 3
1.4 What changes are likely to occur that could alter the frequency and number of toxicity studies submitted? p. 17
2 Introduction on QSARS for Pesticides p. 19
3 Regulatory Perspectives in the use of QSARs p. 22
3.1 Current use of QSARs in regulation p. 22
3.2 Potential barriers for using QSARs in the pesticide authorization procedure p. 27
3.3 End-user criteria for the use of QSARs in regulatory assessment p. 29
4 Quality Criteria for Modelling Ecotoxicity Data p. 30
4.1 Data quality and precision required p. 30
4.2 Quality criteria to be applied to ecotoxicity data used in a QSAR p. 30
4.3 Degree of precision required of QSARs for pesticide assessments p. 39
5 Toxicity End-Points with a High Potential to be Replaced with a QSAR Approach p. 47
5.1 Data availability p. 47
5.2 Number of animals tested p. 48
5.3 Study costs p. 50
5.4 End-points with high potential for replacement with a QSAR p. 51
5.5 Priority end-points p. 54
References p. 54
Chapter 2 Databases for pesticide ecotoxicity Emilio Benfenati and Elena Boriani and Marian Craciun and Ladan Malazizi and Daniel Neagu and Alessandra Roncaglioni p. 59
1 Introduction p. 59
2 Data Availability p. 60
2.1 The EPA-OPP database p. 61
2.2 The SEEM database p. 62
2.3 The BBA database p. 64
2.4 Other databases p. 64
3 Selection of the Data p. 65
3.1 Key features in the choice of the database p. 66
3.2 Comparison of the data internally to the database p. 67
4 Data Representation for Predictive Toxicology p. 70
4.1 A public database example: DSSTox p. 72
4.2 Current toxicity database limitations p. 72
4.3 XML-based standards in chemistry and toxicology p. 73
4.4 PToxML - a simple XML-based description in predictive toxicology p. 73
5 The Characteristics of the Final Data Sets p. 78
6 Conclusions p. 78
Acknowledgments p. 80
References p. 80
Chapter 3 Characterization of chemical structures Emilio Benfenati and Mose Casalegno and Jane Cotterill and Nick Price and Morena Spreafico and Andrey Toropov p. 83
1 Introduction p. 83
2 Characterization of Bi-dimensional Structures p. 85
2.1 Preprocessing of compounds in the data set p. 86
2.2 Geometrical isomers p. 88
2.3 Tautomers p. 88
3 Characterization of Tri-dimensional Structures p. 89
3.1 Crystallographic data p. 89
3.2 Conformational searching and energy minimization p. 90
3.3 Stereoisomers p. 93
3.4 Procedure for the quality control of the chemicals and chemical structures p. 94
4 Chemical Structure File Formats p. 94
4.1 Bi-dimensional descriptors p. 95
4.2 Tri-dimensional descriptors p. 99
4.3 Fragments and Residues in DEMETRA p. 102
References p. 107
Chapter 4 Algorithms for (Q)SAR model building Qasim Chaudhry and Jacques Chretien and Marian Craciun and Gongde Guo and Frank Lemke and Johann-Adolf Muller and Daniel Neagu and Nadege Piclin and Marco Pintore and Paul Trundle p. 111
1 Introduction p. 111
2 Methods for Data Pre-Processing and Selecting Descriptors p. 112
3 Models with Classifiers p. 114
3.1 FISs p. 114
3.2 Adaptive fuzzy partition p. 116
3.3 k-NN methods p. 120
4 Models with Regression Systems p. 125
4.1 Traditional linear regression QSAR models p. 125
4.2 ANNs and fuzzy neural networks p. 128
4.3 Self-organizing statistical-learning networks p. 134
5 Conclusions p. 143
References p. 144
Chapter 5 Hybrid systems Nicolas Amaury and Emilio Benfenati and Severin Bumbaru and Antonio Chana and Marian Craciun and Jacques R. Chretien and Giuseppina Gini and Gongde Guo and Frank Lemke and Viorel Minzu and Johann-Adolf Muller and Daniel Neagu and Marco Pintore and Silviu Augustin Stroia and Paul Trundle p. 149
1 Introduction: Goals of the Hybrid Systems p. 149
2 Our Hybrid Approach for Quantitative Structure-Activity Relationship p. 151
3 Gating Networks p. 152
3.1 Introduction p. 152
3.2 Gating networks for predictive toxicology - a new approach based on descriptors clustering p. 154
3.3 Hybrid neural fuzzy systems p. 157
3.4 Gating networks as HISs - a data-driven approach p. 159
4 Multi-Classifier Systems p. 160
4.1 Approaches for multi-classifier systems p. 161
4.2 An architecture of MCS p. 162
4.3 Classifiers p. 163
4.4 Combination Methods p. 163
4.5 Distributed multi-classifier systems p. 165
5 Neural Ik- and Ek-Based Systems - Introduction of the Prototype NIKE p. 167
5.1 Experiment 1 p. 173
5.2 Experiment 2 p. 174
6 Rule-Based Systems p. 175
7 Self-Organizing Statistical Learning Networks p. 177
8 Conclusions p. 180
References p. 180
Chapter 6 Validation of the models Emilio Benfenati and Jacques R. Chretien and Giuseppina Gini and Nadege Piclin and Marco Pintore and Alessandra Roncaglioni p. 185
1 Introduction p. 185
2 Selection of the Training and Test Sets p. 186
3 Internal Validation and Robustness p. 187
4 External Validation p. 189
5 Validation Parameters for Classifiers: Matrix of Confusion p. 191
6 Graphical Evaluation of the Models: The Receiver Operating Characteristic and Regression Error Characteristic Curves p. 192
7 How to Deal with False Negatives/False Positives p. 196
8 The Applicability Domain p. 197
References p. 198
Chapter 7 Results of DEMETRA models Nicolas Amaury and Emilio Benfenati and Elena Boriani and Mose Casalegno and Antonio Chana and Qasim Chaudhry and Jacques R. Chretien and Jane Cotterill and Frank Lemke and Nadege Piclin and Marco Pintore and Chiara Porcelli and Nicholas Price and Alessandra Roncaglioni and Andrey Toropov p. 201
1 Overview of Results with the Regression Approach p. 201
2 Overview of the Prediction Results Obtained by Classification Methods p. 208
2.1 Data sets and toxicity intervals p. 208
2.2 Descriptors selection and classification results p. 208
2.3 Conclusions about classification results p. 214
3 Overview of Results of Local Models p. 215
3.1 Chemical classes p. 215
4 Overview of Results Obtained with the Hybrid Models p. 221
4.1 Hybrid model for rainbow trout p. 222
4.2 Outliers and the applicability domain p. 226
4.3 Hybrid model for water flea (Daphnia magna) p. 249
4.4 Hybrid model for quail: oral exposure p. 266
4.5 Hybrid model for quail: dietary exposure p. 269
4.6 Hybrid model for acute contact toxicity of honey bee p. 274
5 Conclusions p. 279
Acknowledgments p. 280
References p. 281
Chapter 8 The quality criteria of the DEMETRA models for regulatory purposes Emilio Benfenati p. 283
1 The OECD Guidelines for QSAR Models p. 283
1.1 Introduction p. 283
1.2 The identification of the regulation p. 284
1.3 The criteria for the endpoint selection p. 284
1.4 The model utility p. 285
1.5 The endpoint selection: identification of the guidelines p. 285
1.6 The accordance of the toxicity data to the guidelines p. 286
1.7 The check of quality data p. 286
1.8 The definition of the model components. OECD principle number 2: an unambiguous algorithm p. 286
1.9 The selection of the toxicity values of the data set p. 287
1.10 The characterization of the uncertainty of the experimental data p. 287
1.11 The chemical structures p. 288
1.12 The chemical descriptors p. 289
1.13 The algorithms p. 289
1.14 The performances of the model p. 289
1.15 The reproducibility of the models p. 290
1.16 The false-negative issue p. 290
1.17 The applicability domain p. 291
1.18 The quality control p. 292
1.19 The use of the model p. 292
2 The Specificity of the QSAR Models for Regulatory Purposes p. 292
3 The Probabilistic Meaning of the Model, the Prediction of the Effect, and the Prediction of the Mechanism p. 295
3.1 The probabilistic nature of the models p. 295
3.2 The mechanistic basis of the models p. 296
3.3 The final model and the ways to obtain it p. 297
4 The Benefits of the DEMETRA Models p. 297
5 Future Perspectives p. 298
References p. 301
Chapter 9 The use of the DEMETRA models Emilio Benfenati and Marian Craciun and Daniel Neagu p. 303
1 Introduction p. 303
2 The Users of the DEMETRA Models p. 303
3 Ownership of the Software p. 304
4 Using DEMETRA Models p. 306
5 Chemical Restrictions of the DEMETRA Models p. 307
6 The Format for Model Presentation for DEMETRA: HISML p. 308
References p. 312
Appendices p. 315
Appendix A Summary of responses to DEMETRA survey p. 317
Appendix B Toxicity values for five ECOTOX data sets for pesticide p. 323
Appendix C Example procedures in molecular modelling p. 463
Appendix D The descriptors selected for each data set p. 469
Appendix E List of abbreviations p. 487
Appendix F Software tool for toxicity prediction of pesticides, candidate pesticides, and their derivatives (user guide) p. 493
Index p. 505