QSARINS-Chem. Paola Gramatica, Alessandro Sangion, Ester Papa

Transcription

1 Benign by design approach for more sustainable chemicals: prioritization of chemicals of environmental concern by QSAR modelling in QSARIS-Chem Paola Gramatica, Alessandro Sangion, Ester Papa QSAR Research Unit in Environmental Chemistry and Ecotoxicology University of Insubria, Varese (ITALY) ICCE- Oslo (orway) June 2017

2 Principle 4: Design Safer (Benign) Chemicals Wherever practicable, synthetic methodologies should be designed to use and generate substances that possess little or no toxicity to human health and the environment Principle 10: Design for Degradation Chemical products should be designed so that at the end of their function they do not persist in the environment and break down into innocuous degradation products. BEIG by Structural DESIG using QSAR

3 QSAR for synthesis of sustainable chemicals The HAZARD is an inherent property of a chemical A chemical has the potentiality to be dangerous included in its MOLECULAR STRUCTURE EXPLOIT AVAILABLE KOWLEDGE! A recognized harmful behaviour of a chemical must be taken into consideration to avoid similar concerns for new chemicals. From QSAR-based screening: Priority Lists of the most hazardous chemicals can be proposed AVOID SYTHESIS of REGRETTABLE CHEMICALS!

4 Free licence request in P.Gramatica et al. J. Computational Chemistry (Software ews and Updates) 34, , , , 2014.

5 MODELIG in QSARIS: not «PUSH a BUTTO» Molecular structures are drawn and minimized by the semi-empirical method AM1 (HYPERCHEM software, hin files), and converted into SMILES or MOL files (Open Babel) CAS AMW Sv Ss Mv Me Ms Mono-, bi-, three-dimensional descriptors are calculated using PaDEL (free on line) Multiple linear regression (MLR) performed by Ordinary Least Squares (OLS) method Variable selection by Genetic Algorithm (GA) Validation by many statistical parameters. Various plots, also for Applicability Domain.

6 Chemicals of Environmental Concern (CEC) Persistent Organic Pollutants: from 12 «legacy» POPs Persistent BioaccumulativeToxics (PBTs) CADASTER EU-Project Emerging pollutants: Flame Retardants (FRs), Perfluorinated (PFC), Personal Care Products (PCPs) PCPs and Pharmaceuticals, together named PPCPs

7 Screening of POPs and PBTs COMBIED APPROACH: Multivariate Analysis by PCA + QSAR Two PCA were made on: 1. Half-lives in different environmental media for POPs 2. PBT properties PC1 : ew Macro- Variables 1. GLOBAL Half-Life Index GHLI 2. PBT Index QSAR Models MLR OLS in QSARIS Predictive Tools for Screening and Prioritization of potentially high concern chemicals, based on their cumulative properties P. Gramatica Chemometric Methods and Theoretical Molecular Descriptors in Predictive QSAR Modeling of the Environmental Behaviour of Organic Pollutants, Ch. 12 in Recent Advances in QSAR Studies, Springer-Verlag, 2009

8 PC2 (Explained Variance = 16%) Screening of POPs: Global Half Life Index - GHLI P. Gramatica, E. Papa, Environ. Sci & Technol. 41, 28, Region 2 low Global HL medium LRT potential ot-pops not POP P in Air P in 1 medium (not Air) P in 2 media P in 3 media POP Region 4 low Global HL no LRT potential First two PCs explain 94% of total information on Half-lives HL AIR 1,2,3,4,7-PCDD 2,3,7,8 TCDF HL WATER 4 HL SOIL PCB3 GHL-Index HL SEDIMET GHLI Predicted chlordane aldrin PC1 (Explained Variance = 78%) - GHLIndex Region 1 high Global HL high LRT potential PCB209 PCB208 HCB PCB169 PCB173 PCB122 1,2,3,4,6,7,8 HCDD OCDF PCB21 p,p'ddt 3 lindane 2 toxaphene dieldrin Training Prediction reference chemicals p,p'dde o,p'dde POPs Region 3 high Global HL low LRT potential PCB 0 p-cresol n tr. = 125 R 2 = 0.85 Q 2 LOO = 0.83 RMSE=0.76 n pr. = 125 Q 2 EXT = 0.79 RMSEP=0.78 atrazine aldrin PCB28 α-hch PCB101 CCl 4 PCB180 HCB GHLI from PCA QSAR MODEL GHL Index Y= X0v+5.1 Mv-0.32 MAXDP nhdon-0.5 CIC O

9 Screening of PBTs: QSAR model of PBT Index GHLI not PBT PBT BCF Tox. (Pim.prom.) Global Index of Persistence Bioaccumulation and Toxicity PBT-Index QSAR-Model E.Papa, P. Gramatica, GREE CHEMISTRY 2010, 12, Hot Article PBT Index = nX +0.21nBM nhdon MAXDP PBTs and vpvb ntr=94; R2 = 0.88; Q2 = 0.86; RMSE T =0.53; npred=93; Q2ext = 0.87; RMSE P =0.55

10 PBT Assessment and Prioritization by Consensus Modeling 2015, 77, Agreement graph 4412 chemicals screened by Insubria PBT Index and US-EPA PBT Profiler Consensus Approach (>80% agreement) 1313 chemicals identified as PBTs by Consensus: PRIORITY LIST of SVHC 31% of chemicals overestimated by the PBT Index More Precautionary

11 Screening of ew Brominated Flame Retardants (BFRs) and OrganoPhosphorous FR (OPFRs) by PBT Index , 306, 237 Insubria Graph for model AD PBT Index highlights, from the design, that: Some «safer alternatives» are not so safe: BRFs are mainly PBTs OPFRs are safer (but probably not all ) Pred. by model eq PBT Training set ot PBT Medium PBT PBT OPFRs ot PBT Medium PBT PBT BFRs ot PBT Medium PBT PBT Applicability Domain (AD) of QSARIS PBT Index model HAT i/i (h*=0.0833) 11

12 PBT behaviour of Pharmaceuticals and PCPs (PPCPs) Comparison and consensus of two different modeling approaches for the PBT assessment Insubria PBT Index - Cumulative index for P, B and T, inherent to chemical structure (QSARIS) US-EPA PBT Profiler - Separate predictions for P, B and T 2015, 1, , 147, > 500 Personal Care Products > 1200 Pharmaceuticals 25

13 Screening of PPCPs for PBT Behaviour all % Agreement Overestimated by Profiler 75 4 Overestimated by Index Consensus approach non-pbts in agreement Interp Extrap All Insubria PBT Index more precautionary US-EPA PBT Profiler In AD Bor AD Out AD Agreement graph Overestimated by Profiler on PBT PBT PBTs in agreement Priority List: 8 PCPs and 35 pharmaceuticals InsubriaPBT Index Overestimated by Index chemicals labelled according to the AD

14 Priority List of PCPs potentially PBTs 8 PCPs were predicted as potential PBT by consensus: they are mainly benzotriazoles, used as UV-filters in Sun- Screeners. Cl Cl HO HO HO H O O OH HO O O OH

15 Aquatic toxicity of PCPs and Pharmaceuticals (PPCPs) The aquatic toxicity of environmental chemicals is studied by toxicity tests on three different trophic levels: algae, daphnids and fish. Very limited experimental data availability for PPCPs, mainly of homogeneous data, usefull for QSAR modeling. Data Curation for homogeneous datasets: same species, time, end point etc.

16 QSAR models for Eco-toxicity of PCPs D A T A C U R A T I O Models summary n R 2 Q 2 LOO Q 2 LMO RMSE cv CCC ext Q 2 ext Fn %AD pec50 P. subcapitata pec50 D. magna plc50 P. promelas Eq. Stat. Par. Eq. Stat. Par. Eq. Stat. P. Gramatica, S. Cassani, A. Sangion 2016, 18, Par GGI Mp % MW -3.21ATSC0c 1.52GATS1p % XlogP+2.356minHother AVP % Externally Predictive Models High AD to more than 500 PCPs 16 30

17 QSAR models for Eco-toxicity of Pharm. D A T A C U R A T I O Models summary n R 2 Q 2 LOO Q 2 LMO RMSE cv CCC ext Q 2 ext Fn %AD pec50 P. subcapitata pec50 D. magna plc50 O. mykiss Eq minHother -4.61VCH piPC VE3_Dt Stat. Par Eq % CrippenLogP minhbint SpMAD_Dzs 0.75 AATSC4i C2SP3 Stat. Par Eq ZMIC2-0.14maxHBint2-1.15HybRatio +0.07AATSC0v Stat. Par % 96% plc50 P. promelas Eq Kier AATS3v 0.39nHBAcc+1.11SpMin7_Bhp Stat. Par , 95, % AD to 1200 Pharm

18 Aquatic toxicity trend of PCPs and Pharm. Combining toxicities on different organisms: overall aquatic toxicity Aquatic Toxicity Index by PCA: ATI Green Chem., 2016, 18, 4315 Environ. Int., 2016, 95, 131 Plots for QSARs and AD

19 All the Insubria MLR QSAR models are available and applicable in QSARIS- Chem Poster 94 in Environmental Modelling Session Predictive QSAR models and chemometrics are useful for screening/prioritisation of large sets of chemicals, and for the design of alternative/sustainable chemicals P. Gramatica. Prioritization of Chemicals Based on Chemoinformatic Analysis, 2016 Handbook of Computational Chemistry Vol 5, Ed. J. Leszczynski, Springer

20 Acknowledgements These studies have been financially supported by: FP7th-EU Project CADASTER Italian MIUR-PRI07 Insubria University fellowships Thanks for your attention!!

21 Thanks to my collaborators: Ester Papa, Simona Kovarich, Huanxiang Liu, Barun Bhhatarai, Jiazhong Li, Mara Luini, icola Chirico, Stefano Cassani, Partha Pratim Roy, Alessandro Sangion