QSPR modeling aqueous solubility of polychlorinated ...E.A. Castro et al. / Central European Journal of Chemistry 2(3) 2004 500 523 509 Probe 1 Probe 2 Probe 3 Correlation weights

CEJC 2(3) 2004 500ndash523

QSPR modeling aqueous solubility of polychlorinatedbiphenyls by optimization of correlation weights of

local and global graph invariants

Eduardo A Castro1lowast Andrey A Toropov2 Alexandra I Nesterova2Ozad M Nabiev2

1 INIFTA Departamento de QuımicaFacultad de Ciencias Exactas Universidad Nacional de La Plata

Diag 113 y 64 Suc 4 CC 16 La Plata 1900 Argentina2 Algorithm-Engineering InstituteUzbekistan Academy of Sciences

F Khodjaev Street 25 700125 Tashkent Uzbekistan

Received 27 October 2003 accepted 25 May 2004

Abstract Aqueous solubilities of polychlorinated biphenyls have been correlated withtopological molecular descriptors which are functions of local and global invariants of labeledhydrogen filled graphs Morgan extended connectivity and nearest neighboring codes have beenused as local graph invariants The number of chlorine atoms in biphenyls has been employedas a global graph invariant Present results show that taking into account correlation weightsof global invariants gives quite reasonable improvement of statistical characteristics for theprediction of aqueous solubilities of polychlorinated biphenylscdeg Central European Science Journals All rights reserved

Keywords Aqueous solubility polychlorinated biphenyls QSPR modeling graph invariantstopological descriptors

1 Introduction

Since the atmosphere is a significant pathway for the transport of organic pollutants

considerable efforts have been expended for the measurement of physicochemical

properties that govern the movement of chemicals in the environment Polychlorinated

biphenyls (PCBs) a class of persistent organic chemicals have attracted the attention of

scientists in recent decades because they are found at an appreciable concentration in the

lowast E-mail castroquimicaunlpeduar jubertarnetcomarUnauthenticated

Download Date | 82319 947 PM

brought to you by COREView metadata citation and similar papers at coreacuk

provided by Servicio de Difusioacuten de la Creacioacuten Intelectual

EA Castro et al Central European Journal of Chemistry 2(3) 2004 500ndash523 501

polar regions presumably as a result of long-range atmospheric transport [1] Although

the manufacture and use of PCBs have been banned since 1979 [2] these persistent

organic pollutants remain widely distributed in the environment due to their chemical

stability Among the environmental pollutants that may be able to disrupt the endocrine

system of human and animals PCBs have been particularly noteworthy [34]

The ability of PCBs to mimic natural hormones may reflect a close relationship

between the physicochemical properties encoded in the molecular structure of these

compounds and the toxic responses they elicit in biological systems Due to their

remarkable insulating capacity and flame resistant nature PCBs replaced combustible

insulating fluids in capacitors and transformers and reduced the risk of fire in hospitals

schools and factories PCBs entered the environment as components of pesticides

plasticizers and adhesives The nonflammability and chemical stability of PCBs have

contributed to the widespread environmental problems associated with these

organohalogen compounds The lipophilicity of these compounds is responsible for

their accumulation in the food chain and the cause of adverse human health effects

In light of the probable carcinogenic activity of these compounds [56] and their

tendency to be sorbed and bio-accumulated in aquatic environments aqueous solubility

(Sw) of the PCBs has been measured by a variety of investigators [7-11] Recently Puri

et al [12] calculated Sw of PCBs using Mobile Order and Disorder Theory [7-8] for a

representative set of 61 molecules They obtained good agreement between calculated

and experimental solubility values of PCBs at 29815 K (standard deviation = plusmn041 log

units Table 6 in Ref 12) which demonstrates the utility and capability of Comparative

Molecular Field Analysis (CoMFA)-predicted values of fusion enthalpies to calculate the

Sw of any PCB The aim of CoMFA [13-15] is to derive a correlation between the biological

activity of a series of molecules and their 3D shape electrostatic and hydrogen bonding

characteristics

Since there are other alternatives to predict Sw within the frame of the QSPR theory

we have looked for ways to improve these predictions An option is the approach based

upon correlation weights of local graph invariants [16-19] which has proved to be a

suitable tool to calculate thermodynamic properties for a wide variety of molecular species

[20-24]

The aim of the present study was to develop simple and predictive models that

correlate Sw of PCBs with Morgan extended connectivity and nearest neighboring codes

This study shows that this particular set of molecular topological descriptors makes up

a suitable option to predict this physicochemistry property with a greater accuracy than

previous approaches so that these models provide a numerical value that can be used in

cases when experimental data are unavailable

Since the atmosphere is a significant pathway for the transport of organic pollutants

considerable efforts have been expended for the measurement of physicochemical

properties that govern the movement of chemicals in the environment

UnauthenticatedDownload Date | 82319 947 PM

502 EA Castro et al Central European Journal of Chemistry 2(3) 2004 500ndash523

2 Method

The modeling of the Sw of PCBs was based on the optimization of the correlation weights

of graph invariants in the Labeled Hydrogen-Filled Graph (LHFG) and in the Graph

of Atomic Orbitals (GAO) versions Since the methodological principles and specific

formulae have been presented elsewhere [16-24] it is not necessary to introduce them

again

The molecular descriptors are defined as

D(a LI) =Xk

CW (ak) +Xk

CW (LIk) (1)

D(ao LI) =Xk

CW (aok) +Xk

CW (LIk) (2)

where CW is the correlation weight ak is a chemical element that is image of the k-th

vertex in the LHFG aok is the atomic orbital that is the image of the k-th vertex in the

GAO LIk is some numerical local invariant of LHFG or GAO As local invariants (ieLIs)

we have chosen the Morgan extended connectivity of zero (0EC) and first (1EC) order in

the LHFG and also in the GAO nearest neighboring codes (NNC) in the LHFG

The Sw of 61 PCBs have been reported in the literature [7 8] This molecular set has

been chosen by Puri et al [12] to develop three-dimensional quantitative-structure-property

relationship (3D-QSPR) models for prediction of enthalpies of fusion and their application

to estimates of enthalpies of sublimation and Sw and we have selected this set of PCBs

to be able to make a direct comparison of our predictions with previous results

We have chosen two calculation strategies to report results

(a) We have made calculations on the whole molecular set of PCBs (ie 61 molecules)

(b) We have divided the complete molecular set into two partial sets a training set (31

molecules) and a test set (30 molecules) The regression models were determined

according to the training set and true predictions were made for the molecules

belonging to the test set

Since in principle the partition is arbitrary we have tried several choices in order to

determine the dependence of final results on such partitions However we have found

that final results are nearly independent of the chosen partition so that we report results

for a typical choice Tables 8-12 list the composition of each partial sets

3 Results and discussion

As it is usual in these kind of calculations we have tested more than one numerical probe

In order to reach internal consistency we have tried three different probes to test some

possible dependencies on a particular one

In Table 1 we present results of OCWLI based on local graph invariants in the

LHFGs and in the GAO From the results shown in Table 1 one can see that results

are independent of the probes for each variable so that they are internally consistent

Statistical characteristics are nearly the same in the five cases presented here Such aUnauthenticatedDownload Date | 82319 947 PM

EA Castro et al Central European Journal of Chemistry 2(3) 2004 500ndash523 503

situation is most likely due to the great similarity among molecular structures under

consideration In other words descriptors calculated with the CWs often have equal

numerical values Under such circumstances taking into account the global

graph invariants becomes a reasonable concept of modeling for this molecular set The

optimization of correlation weights of the mentioned local invariants together with the

number of chlorine atoms that are present in the LHFG of a PCB (denoted as NCl) and

the number of 3p5 orbitals which are present in the GAO of a PCB (denoted as N3p5)

may be considered one of the possible ways of defining local and global optimization

scheme in the QSPR modeling In other words the QSPR analysis of PCBs descriptors

are calculated as

D(a LI) =

(Xk

CW (ak) +Xk

CW (LIk)

)+ CW (NCl) (3)

D(ao LI) =

(Xk

CW (aok) +Xk

CW (LIk)

)+ CW (N3p5) (4)

Results derived from the calculation with Eqs (3)-(4) are presented in Table 2 One can

see that statistical characteristics of models displayed in Table 2 are better than those

ones given in Table 1 Final results are also nearly independent of the chosen probe as

seen with data analyzed in Table 1

Correlation weights for calculating D(aLI) of Eqs (3)-(4) are presented in Tables 3-7

The results derived from the calculation the Sw of the PCBs with the optimized fitting

linear polynomials are shown in Tables 8-12

LogSw = 2599D(a 0EC)minus 3148 (5)

LogSw = 1870D(a 1EC)minus 1275 (6)

LogSw = 05388D(ao 0EC)minus 7915 (7)

LogSw = 03547D(ao 1EC)minus 7896 (8)

LogSw = 1659D(a NNC)minus 9758 (9)

The analysis of data shows satisfactory agreement among experimental and theoretical

predictions of Sw Particularly notable are the lower absolute average deviations for the

test set save the predictions derived from Eq (11) (033 vs 040) of Ref 12 In fact

although differences are not spectacular (ie 038 vs 035 (twice) 038 vs 033 and 039

vs 037) they are significant Large deviations are scarce and they amount to around

16 (for example molecules 14 and 15 in Tables 8 9 and 12 (training set) molecule 12

in Table 11 (test set)) Once again we found a good predictive capability in the fitting

equations since in one case (molecule 12 Table 11) the deviation is rather large for a

member of the test set In order to judge the suitably of these findings we must take into

account that results for test sets are true predictions and not the outcome of numerical

fittings

The comparison with previous results [12] for this molecular set shows the relative

merits of the present approach In fact the average absolute deviation obtained by PuriUnauthenticatedDownload Date | 82319 947 PM

504 EA Castro et al Central European Journal of Chemistry 2(3) 2004 500ndash523

et al was 032 (see Table 6 in Ref 12) which is smaller than the present results However

since the authors did not divide the total molecular set into a training and a test set the

solubilities calculated from their calculation of fusion enthalpies are not true predictions

as are the results of this work

4 Conclusion

The optimization of correlation weights of local and global graph invariants in the LHFG

andor GAO approaches may be considered a reasonably good tool to predict the Sw of

the PCBs for the molecular set under consideration in the present study The relative

deviations are relatively low and true predictions are satisfactory In fact from a general

viewpoint the average absolute deviations are rather small and in only one case have

we found a relatively large deviation for a member of the test set We conclude that

the present approach based on the optimization of correlation weights of local and global

graph invariants is a suitable way of predicting Sw of PCBs This finding is in line with our

previous findings about these special kinds of molecular descriptors Global descriptors

appear to be better variables than local ones in calculating Sw of this particular set of

PCBs

Acknowledgment

Authors thanks valuable comments made by two anonymous referees which have been

very helpful to improve the final version of this article

References

[1] F Wania and D Mackay ldquoGlobal Fractionation and Cold Condensation of LowVolatility Organochlorine Compounds in Polar Regionsrdquo AMBIO Vol 22 (1993)pp 10ndash18

[2] Agency for Toxic Substances and Disease Registry (ATSDR) 1993 ToxicologicalProfile for Selected PCBs Atlanta US Public Health Service US Department ofHealth and Human Services

[3] JD McKinney and CL Waller ldquoPolychlorinated Biphenyls as Hormonally ActiveStructural Analoguesrdquo Environ Health Perspect Vol 102 (1994) pp 290ndash297

[4] A Brouwer MP Longnecker LS Birnbaum J Cogliano P Kostyniak J MooreS Schantz and G Winneke ldquoCharacterization of Potential Endocrine-Related HealthEffects at Low-Dose Levels of Expopsure to PCBsldquo Environ Health Perspect Vol107 (1999) pp 639ndash649

[5] US Environmental Protection Agency 1994 Integrated Risk Information System(IRIS) on PCBs Cincinnati OH Environmental Criteria and Assesment OfficeOffice of Health and Environmental Assessment Office of Research and Development

[6] US Environmental Protection Agency 1996 PCBs Cancer Dose-ResponseAssessment and Application to Environmental Mixtures Report EPA 600P-96002FUnauthenticated

Download Date | 82319 947 PM

EA Castro et al Central European Journal of Chemistry 2(3) 2004 500ndash523 505

[7] P Ruelle and UW Kesserling ldquoAqueous Solubility Prediction of EnvironmentallyImportant Chemicals from the Mobile Order Thermodynamicrdquo Chemosphere Vol34 (1997) pp 275ndash298

[8] P Ruelle A Cuendet and UW Kesselring ldquoHydrophobic and Solvation Effects onthe Solubility of Hydroxysteroids in Various Solvents Quantitative and QualitativeAssessment by Application of the Mobile Order and Disorder Theoryrdquo PerspectiveDrug Discovery Design Vol 18 (2000) pp 61ndash112 and references herein

[9] WJ Doucette and AW Andren ldquoEstimation of OctanolWater PartitionCoefficients Evaluation of Six Methods for Highly Hydrophobic AromaticHydrocarbonsldquo Chemosphere Vol 17 (1988) pp 345ndash359

[10] R Kuhne R-U Ebert F Kleint G Schmidt and G Schuurmann ldquoGroupContribution Methods to Estimate Water Solubility of Organic ChemicalsrdquoChemosphere Vol30 (1995) pp 2061ndash2077

[11] TT Blair E Gifford WE Acree Jr and C-C Tsai ldquoQuantitativeStructure-Property Relationships for Aqueous Solubilities of Halogenated AromaticCompoundsrdquo Phys Chem Liq Vol 24 (1992) pp 137ndash160

[12] S Puri JS Chickos and WJ Welsh ldquoThree-Dimensional Quantitative Structurendash Property Relationship (3D-QSPR) Models for Prediction of ThermodynamicProperties of Polychlorinated Biphenyls (PCBs) Enthalpies of Fusion and TheirApplication to Estimates of Enthalpies of Sublimation and Aqueous Solubilitiesrdquo JChem Inf Comput Sci Vol 43 (2003) pp 55ndash62

[13] RD Cramer III DE Patterson and JE Bunce ldquoComparative Molecular FieldAnalysis (CoMFA) 1 Effect of Shape on Binding of Steroids to Carriers ProteinsrdquoJ Am Chem Soc Vol 110 (1988) pp 5959ndash5967

[14] G Klebe ldquoStructural Alignment of Moleculesldquo In H Kubinyi(Ed) 3D-QSAR inDrug Design ESCOM Leiden 1993 pp 173ndash199

[15] TI Oprea and CL Waller ldquoTheoretical and Practical Aspects of Three DimensionalQuantitative Structure-Activity Relationshipsldquo In KB Lipkowitz and DBBoyd(Eds) Reviews in Computational Chemistry Vol 11 Wiley New York 1997pp 127ndash182

[16] PJ Peruzzo DJG Marino EA Castro and AA Toropov ldquoCalculation of pKValues of Flavylium Salts from the Optimization of Correlation Weights of LocalGraph Invariantsrdquo J Mol Struct THEOCHEM Vol 572 (2001) pp 53ndash60

[17] DJG Marino PJ Peruzzo EA Castro and AA Toropov ldquoQSAR CarcinogenicStudy of Methylated Polycyclic Aromatic Hydrocarbons Based on TopologicalDescriptors Derived from Distance Matrices and Correlation Weights of LocalGraph Invariantsrdquo Internet Electron J Mol Des Vol 1 (2002) pp 108ndash133httpwwwbiochempresscom

[18] AA Toropov and AP Toropova ldquoOptimization of Correlation Weights of the LocalGraph Invariants Use of the Enthalpies of Formation of Complex Compounds forthe QSPR Modelingrdquo Russ J Coord Chem Vol 24 (1998) pp 81ndash85

[19] AA Toropov and AP Toropova ldquoModeling of Acyclic Compounds Normal BoilingPoints by Correlation Weighting of Nearest Neighboring Codesrdquo J Mol StructTHEOCHEM Vol 581 (2002) pp 11ndash15

[20] AA Toropov OM Nabiev PR Duchowicz EA Castro and F Torrens ldquoQSPRModeling of Hydrocarbon Dipole Moments by Means of Correlation Weighting ofLocal Graph Invariantsrdquo J Theor Comp Chem Vol 2 (2003) pp 139ndash146Unauthenticated

Download Date | 82319 947 PM

506 EA Castro et al Central European Journal of Chemistry 2(3) 2004 500ndash523

[21] DJG Marino PJ Peruzzo EA Castro and AA Toropov ldquoQSPR Modeling ofLipophilicity by Means of Correlation Weights of Local Graph Invariantsrdquo InternetElect J Molec Design Vol 2 (2003) pp 334ndash347

[22] AA Toropov PR Duchowicz and EA Castro ldquoStructure-Toxicity Relationshipsfor Aliphatic Compounds Based on Correlation Weighting of Local Graph InvariantsrdquoInt J Mol Sci Vol 2 (2003) pp 272ndash283

[23] PR Duchowicz EA Castro and AA Toropov ldquoQSPR Modeling of Normal BoilingPoints of Aldehydes Ketones and Esters by Means of Nearest Neighboring CodesCorrelation Weightingrdquo J Arg Chem Soc Vol 90 (2002) pp 91ndash107

[24] PR Duchowicz EA Castro and AA Toropov ldquoImproved QSPR Analysis ofStandard Entropy of Acyclic and Aromatic Compounds Using Optimized CorrelationWeights of Linear Graph Invariantsrdquo Comp Chem Vol 26 (2002) pp 327ndash332

UnauthenticatedDownload Date | 82319 947 PM

EA Castro et al Central European Journal of Chemistry 2(3) 2004 500ndash523 507

Training Set (n = 31) Test Set (n = 30) Complete set (n = 61)Descriptor r s F r s F r s F

Probe 1 D(a0EC) 09143 0609 148 09623 0404 351 09320 0514 390Probe 2 D(a0EC) 09143 0609 148 09623 0404 351 09320 0514 390Probe 3 D(a0EC) 09143 0609 148 09623 0404 351 09320 0514 390Probe 1 D(a1EC) 09143 0609 148 09623 0404 351 09320 0514 390Probe 2 D(a1EC) 09143 0609 148 09623 0404 351 09320 0514 390Probe 3 D(a1EC) 09143 0609 148 09623 0404 351 09320 0514 390Probe 1 D(ao0EC) 09140 0609 148 09620 0404 351 09320 0514 390Probe 2 D(ao0EC) 09140 0609 148 09620 0404 351 09320 0514 390Probe 3 D(ao0EC) 09140 0609 148 09620 0404 351 09320 0514 390Probe 1 D(ao1EC) 09301 0552 186 09328 0477 187 09306 0512 382Probe 2 D(ao1EC) 09301 0552 186 09330 0477 188 09307 0512 382Probe 3 D(ao1EC) 09301 0552 186 09331 0476 189 09308 0512 383Probe 1 D(aNNC) 09143 0609 148 09623 0404 351 09320 0514 390Probe 2 D(aNNC) 09143 0609 148 09623 0404 351 09320 0514 390Probe 3 D(aNNC) 09143 0609 148 09623 0404 351 09320 0514 390

r is the linear correlation coefficients is the standard error of estimatesF is the Fischer ratio

Table 1 Statistical characteristics of PCB solubility models based on local graph invariants

UnauthenticatedDownload Date | 82319 947 PM

508 EA Castro et al Central European Journal of Chemistry 2(3) 2004 500ndash523

Training Set (n = 31) Test Set (n = 30) Complete set (n = 61)Descriptor r s F r s F r s F

Probe 1 D(a0EC) 09372 0525 209 09620 0437 347 09429 0479 473Probe 2 D(a0EC) 09371 0525 209 09626 0437 353 09430 0479 474Probe 3 D(a0EC) 09372 0525 209 09627 0436 354 09431 0479 474Probe 1 D(a1EC) 09372 0524 209 09629 0434 356 09432 0478 476Probe 2 D(a1EC) 09372 0525 209 09619 0434 347 09430 0478 474Probe 3 D(a1EC) 09372 0525 209 09631 0434 358 09432 0478 476Probe 1 D(ao0EC) 09372 0524 209 09627 0434 354 09432 0478 475Probe 2 D(ao0EC) 09372 0524 209 09625 0435 353 09431 0479 474Probe 3 D(ao0EC) 09372 0524 209 09626 0435 354 09431 0478 475Probe 1 D(ao1EC) 09533 0454 289 09141 0524 142 09374 0486 428Probe 2 D(ao1EC) 09534 0454 290 09140 0524 142 09375 0485 428Probe 3 D(ao1EC) 09536 0453 291 09136 0524 141 09375 0485 428Probe 1 D(aNNC) 09372 0524 209 09627 0433 355 09433 0478 476Probe 2 D(aNNC) 09371 0525 209 09621 0432 349 09432 0477 475Probe 3 D(aNNC) 09371 0525 209 09619 0435 347 09430 0479 473

Table 2 Statistical characteristics of PCB solubility models based on local and global graphinvariants

UnauthenticatedDownload Date | 82319 947 PM

EA Castro et al Central European Journal of Chemistry 2(3) 2004 500ndash523 509

Probe 1 Probe 2 Probe 3

Correlation weights of the ak values

H 1140 1037 1140

C 3583 1517 1191

Cl 0898 0889 0934

Correlation weights of the 0EC values

0001 2986 1505 1235

0003 2837 1948 1384

Correlation weights of the NCl values

H000 1173 1287 1221

H001 1070 1199 1113

H002 1002 1128 1045

H003 1080 1150 1099

H004 0963 1046 0988

H005 1210 1200 1198

H006 1128 1113 1105

H007 1287 1195 1230

H008 1244 1150 1188

H009 1002 0950 0965

H010 0777 0769 0750

Table 3 Correlation weights for the calculation of D(a0EC)

UnauthenticatedDownload Date | 82319 947 PM

510 EA Castro et al Central European Journal of Chemistry 2(3) 2004 500ndash523

Probe 1 Probe 2 Probe 3

Correlation weights of the ak values

H 1353 1100 1200

C 1517 1026 0866

Cl 1015 0923 0834

Correlation weights of the 1EC values

0003 1655 1128 1339

0007 1315 1200 1298

0009 1635 0900 1187

Correlation weights of the NCl values

H000 1187 1098 1304

H001 1040 1019 1150

H002 0950 0963 1052

H003 1058 1019 1156

H004 0909 0923 0998

H005 1244 1100 1353

H006 1126 1037 1240

H007 1344 1141 1458

H008 1295 1120 1421

H009 0968 0923 1049

H010 0656 0750 0710

Table 4 Correlation weights for the calculation of D(a1EC)

UnauthenticatedDownload Date | 82319 947 PM

EA Castro et al Central European Journal of Chemistry 2(3) 2004 500ndash523 511

Probe 1 Probe 2 Probe 3

Correlation weights of the aok values

1s1 2022 2048 2457

1s2 0616 0650 0834

2p2 1562 1307 2416

2s2 0649 0720 0569

2p6 0686 0804 0615

3s2 0723 0703 0689

3p5 0620 0745 0554

Correlation weights of the 0EC values

0003 0656 0673 0590

0007 2281 2488 2125

0009 1380 2092 1501

0011 0568 0731 0598

Correlation weights of the N3p5 values

H000 1497 1517 1615

H001 1102 1149 1163

H002 0836 0920 0860

H003 1278 1262 1360

H004 0818 0857 0849

H005 2053 1863 2252

H006 1736 1588 1912

H007 2548 2238 2838

H008 2454 2139 2737

H009 1372 1230 1497

H010 0368 0363 0368

Table 5 Correlation weights for the calculation of D(ao0EC)

UnauthenticatedDownload Date | 82319 947 PM

512 EA Castro et al Central European Journal of Chemistry 2(3) 2004 500ndash523

Probe 1 Probe 2 Probe 3

Correlation weights of the aok values

1s1 3038 2194 3452

1s2 1170 1192 0600

2p2 4861 7942 15895

2s2 0781 1073 1171

2p6 0773 0839 0714

3s2 0711 0703 0692

3p5 0813 0579 0885

Correlation weights of the 1EC values

0021 2790 2828 2757

0033 0604 0651 0769

0045 2372 2893 2561

0051 1596 1945 1755

0057 1893 2188 1944

0063 1302 1467 1381

0069 1360 1410 1275

0075 0946 0909 0912

0081 1069 0951 0930

0093 0804 0530 0617

Correlation weights of the N3p5 values

H000 0459 0318 0424

H001 0812 0723 1001

H002 0785 0750 1032

H003 1469 1696 1976

H004 1349 1561 1836

H005 2940 3660 3834

H006 3006 3823 4008

H007 5072 6495 6688

H008 4633 5961 6215

H009 2246 2880 3039

H010 0068 0018 0028

Table 6 Correlation weights for the calculation of D(ao1EC)

UnauthenticatedDownload Date | 82319 947 PM

EA Castro et al Central European Journal of Chemistry 2(3) 2004 500ndash523 513

Probe 1 Probe 2 Probe 3

Correlation weights of the ak values

H 1440 1230 0975

C 1076 0974 1113

Cl 0953 1128 1100

Correlation weights of the NNC values

0110 1353 0926 0929

0320 1188 1025 0988

0321 1067 1051 1200

0330 1663 0760 1051

Correlation weights of the NCl values

H000 1368 0963 0997

H001 1197 0926 0974

H002 1073 0900 0950

H003 1180 0950 0975

H004 0996 0900 0938

H005 1351 1025 1025

H006 1218 0999 1000

H007 1443 1073 1050

H008 1379 1071 1037

H009 0980 0964 0950

H010 0623 0855 0878

Table 7 Correlation weights for the calculation of D(aNNC)

UnauthenticatedDownload Date | 82319 947 PM

514 EA Castro et al Central European Journal of Chemistry 2(3) 2004 500ndash523

n() Molecule D(a0EC) log Sw Eq (5) Residue

- biphenyl 119473 -431 -429 -0023 4-monochlorobiphenyl 119128 -520 -519 -0018 24rsquo-dichlorobiphenyl 118818 -528 -599 07111 33rsquo-dichlorobiphenyl 118818 -580 -599 01912 34-dichlorobiphenyl 118818 -639 -599 -04015 44rsquo-dichlorobiphenyl 118818 -656 -599 -05718 22rsquo5-trichlorobiphenyl 118654 -602 -642 04031 24rsquo5-trichlorobiphenyl 118654 -625 -642 01737 344rsquo-trichlorobiphenyl 118654 -706 -642 -06444 22rsquo35rsquo-tetrachlorobiphenyl 118295 -647 -735 08852 22rsquo55rsquo-tetrachlorobiphenyl 118295 -700 -735 03566 23rsquo44rsquo-tetrachlorobiphenyl 118295 -668 -735 06775 244rsquo6-tetrachlorobiphenyl 118295 -694 -735 04177 33rsquo44rsquo-tetrachlorobiphenyl 118295 -853 -735 -11880 33rsquo55rsquo-tetrachlorobiphenyl 118295 -854 -735 -11983 22rsquo33rsquo5-pentachlorobiphenyl 118300 -696 -734 03886 22rsquo345-pentachlorobiphenyl 118300 -721 -734 01387 22rsquo345rsquo-pentachlorobiphenyl 118300 -791 -734 -05788 22rsquo346-pentachlorobiphenyl 118300 -743 -734 -009101 22rsquo455rsquo-pentachlorobiphenyl 118300 -733 -734 001104 22rsquo466rsquo-pentachlorobiphenyl 118300 -732 -734 002118 23rsquo44rsquo5-pentachlorobiphenyl 118300 -739 -734 -005128 22rsquo33rsquo44rsquo-hexachlorobiphenyl 117976 -901 -818 -083129 22rsquo33rsquo45-hexachlorobiphenyl 117976 -807 -818 011138 22rsquo344rsquo5rsquo-hexachlorobiphenyl 117976 -832 -818 -014151 22rsquo355rsquo6-hexachlorobiphenyl 117976 -742 -818 076183 22rsquo344rsquo5rsquo6-heptachlorobiphenyl 117893 -792 -840 048187 22rsquo34rsquo55rsquo6-heptachlorobiphenyl 117893 -894 -840 -054202 22rsquo33rsquo55rsquo66rsquo-octachlorobiphenyl 117608 -915 -914 -001208 22rsquo33rsquo455rsquo66rsquo-nonachlorobiphenyl 117124 -1041 -1040 -001209 decachlorobiphenyl 116657 -1162 -1161 -001

Average absolute deviation = 038() IUPAC no

Table 8a Modeling of the PCBs solubility with Eq (5) based on D(a0EC) - training set

UnauthenticatedDownload Date | 82319 947 PM

EA Castro et al Central European Journal of Chemistry 2(3) 2004 500ndash523 515

n Molecule D(a0EC) log Sw Eq (5) Residue

1 2-monochlorobiphenyl 119128 -454 -519 0654 22rsquo-dichlorobiphenyl 118818 -527 -599 07210 26-dichlorobiphenyl 118818 -521 -599 07822 234rsquo-trichlorobiphenyl 118654 -626 -642 01624 236-trichlorobiphenyl 118654 -629 -642 01326 23rsquo5-trichlorobiphenyl 118654 -601 -642 04128 244rsquo-trichlorobiphenyl 118654 -621 -642 02129 245-trichlorobiphenyl 118654 -627 -642 01530 246-trichlorobiphenyl 118654 -614 -642 02833 2rsquo34-trichlorobiphenyl 118654 -629 -642 01340 22rsquo33rsquo-tetrachlorobiphenyl 118295 -728 -735 00747 22rsquo44rsquo-tetrachlorobiphenyl 118295 -651 -735 08449 22rsquo45rsquo-tetrachlorobiphenyl 118295 -657 -735 07853 22rsquo56rsquo-tetrachlorobiphenyl 118295 -708 -735 02754 22rsquo66rsquo-tetrachlorobiphenyl 118295 -721 -735 01461 2345-tetrachlorobiphenyl 118295 -716 -735 01970 23rsquo4rsquo5-tetrachlorobiphenyl 118295 -725 -735 01082 22rsquo33rsquo4-pentachlorobipheny 118300 -705 -734 029116 23456-pentachlorobiphenyl 118300 -792 -734 -058134 22rsquo33rsquo56-hexachlorobiphenyl 117976 -860 -818 -042136 22rsquo33rsquo66rsquo-hexachlorobiphenyl 117976 -865 -818 -047141 22rsquo3455rsquo-hexachlorobiphenyl 117976 -768 -818 050153 22rsquo44rsquo55rsquo-hexachlorobiphenyl 117976 -856 -818 -038155 22rsquo44rsquo66rsquo-hexachlorobiphenyl 117976 -871 -818 -053156 233rsquo44rsquo5-hexachlorobiphenyl 117976 -782 -818 036158 233rsquo44rsquo6-hexachlorobiphenyl 117976 -766 -818 052171 22rsquo33rsquo44rsquo6-heptachlorobiphenyl 117893 -830 -840 010185 22rsquo3455rsquo6-heptachlorobiphenyl 117893 -846 -840 -006194 22rsquo33rsquo44rsquo55rsquo-octachlorobiphenyl 117608 -916 -914 -002206 22rsquo33rsquo44rsquo55rsquo6-nonachlorobiphenyl 117124 -1026 -1040 014

Average absolute deviation = 035

Table 8b Modeling of the PCBs solubility with Eq (5) based on D(a0EC) - test set

UnauthenticatedDownload Date | 82319 947 PM

516 EA Castro et al Central European Journal of Chemistry 2(3) 2004 500ndash523

n Molecule D(a1EC) log Sw Eq (6) Residue

- biphenyl 65891 -431 -428 -0033 4-monochlorobiphenyl 65406 -520 -519 -0018 24rsquo-dichlorobiphenyl 64978 -528 -599 07111 33rsquo-dichlorobiphenyl 64978 -580 -599 01912 34-dichlorobiphenyl 64978 -639 -599 -04015 44rsquo-dichlorobiphenyl 64978 -656 -599 -04318 22rsquo5-trichlorobiphenyl 64748 -602 -642 04031 24rsquo5-trichlorobiphenyl 64748 -625 -642 01737 344rsquo-trichlorobiphenyl 64748 -706 -642 -06444 22rsquo35rsquo-tetrachlorobiphenyl 64261 -647 -733 08652 22rsquo55rsquo-tetrachlorobiphenyl 64261 -700 -733 03366 23rsquo44rsquo-tetrachlorobiphenyl 64261 -668 -733 06575 244rsquo6-tetrachlorobiphenyl 64261 -694 -733 03977 33rsquo44rsquo-tetrachlorobiphenyl 64261 -853 -733 -12080 33rsquo55rsquo-tetrachlorobiphenyl 64261 -854 -733 -12183 22rsquo33rsquo5-pentachlorobiphenyl 64258 -696 -734 03886 22rsquo345-pentachlorobiphenyl 64258 -721 -734 01387 22rsquo345rsquo-pentachlorobiphenyl 64258 -791 -734 -05788 22rsquo346-pentachlorobiphenyl 64258 -743 -734 -009101 22rsquo455rsquo-pentachlorobiphenyl 64258 -733 -734 001104 22rsquo466rsquo-pentachlorobiphenyl 64258 -732 -734 002118 23rsquo44rsquo5-pentachlorobiphenyl 64258 -739 -734 -005128 22rsquo33rsquo44rsquo-hexachlorobiphenyl 63802 -901 -819 -082129 22rsquo33rsquo45-hexachlorobiphenyl 63802 -807 -819 012138 22rsquo344rsquo5rsquo-hexachlorobiphenyl 63802 -832 -819 -013151 22rsquo355rsquo6-hexachlorobiphenyl 63802 -742 -819 077183 22rsquo344rsquo5rsquo6-heptachlorobiphenyl 63682 -792 -842 050187 22rsquo34rsquo55rsquo6-heptachlorobiphenyl 63682 -894 -842 -052202 22rsquo33rsquo55rsquo66rsquo-octachlorobiphenyl 63295 -915 -914 -001208 22rsquo33rsquo455rsquo66rsquo-nonachlorobiphenyl 62630 -1041 -1038 -003209 decachlorobiphenyl 61980 -1162 -1160 -002

Average absolute deviation = 038

Table 9a Modeling of the PCBs solubility with Eq (6) based on D(a1EC) - training set

UnauthenticatedDownload Date | 82319 947 PM

EA Castro et al Central European Journal of Chemistry 2(3) 2004 500ndash523 517

n Molecule D(a1EC) log Sw Eq (6) Residue

1 2-monochlorobiphenyl 65406 -454 -519 0654 22rsquo-dichlorobiphenyl 64978 -527 -599 07210 26-dichlorobiphenyl 64978 -521 -599 07822 234rsquo-trichlorobiphenyl 64748 -626 -642 01624 236-trichlorobiphenyl 64748 -629 -642 01326 23rsquo5-trichlorobiphenyl 64748 -601 -642 04128 244rsquo-trichlorobiphenyl 64748 -621 -642 02129 245-trichlorobiphenyl 64748 -627 -642 01530 246-trichlorobiphenyl 64748 -614 -642 02833 2rsquo34-trichlorobiphenyl 64748 -629 -642 01340 22rsquo33rsquo-tetrachlorobiphenyl 64261 -728 -733 00547 22rsquo44rsquo-tetrachlorobiphenyl 64261 -651 -733 08249 22rsquo45rsquo-tetrachlorobiphenyl 64261 -657 -733 07653 22rsquo56rsquo-tetrachlorobiphenyl 64261 -708 -733 02554 22rsquo66rsquo-tetrachlorobiphenyl 64261 -721 -733 01261 2345-tetrachlorobiphenyl 64261 -716 -733 01770 23rsquo4rsquo5-tetrachlorobiphenyl 64261 -725 -734 00982 22rsquo33rsquo4-pentachlorobiphenyl 64258 -705 -734 029116 23456-pentachlorobiphenyl 64258 -792 -734 -058134 22rsquo33rsquo56-hexachlorobiphenyl 63802 -860 -819 -041136 22rsquo33rsquo66rsquo-hexachlorobiphenyl 63802 -865 -819 -046141 22rsquo3455rsquo-hexachlorobiphenyl 63802 -768 -819 051153 22rsquo44rsquo55rsquo-hexachlorobiphenyl 63802 -856 -819 -037155 22rsquo44rsquo66rsquo-hexachlorobiphenyl 63802 -871 -819 -052156 233rsquo44rsquo5-hexachlorobiphenyl 63802 -782 -819 037158 233rsquo44rsquo6-hexachlorobiphenyl 63802 -766 -819 053171 22rsquo33rsquo44rsquo6-heptachlorobiphenyl 63682 -830 -842 012185 22rsquo3455rsquo6-heptachlorobiphenyl 63682 -846 -842 -004194 22rsquo33rsquo44rsquo55rsquo-octachlorobiphenyl 63295 -916 -914 -002206 22rsquo33rsquo44rsquo55rsquo6-nonachlorobiphenyl 62630 -1026 -1038 012

Average absolute deviation = 033

Table 9b Modeling of the PCBs solubility with Eq (6) based on D(a1EC) - test set

UnauthenticatedDownload Date | 82319 947 PM

518 EA Castro et al Central European Journal of Chemistry 2(3) 2004 500ndash523

n Molecule D(ao0EC) log Sw Eq (7) Residue

- biphenyl 138911 -431 -431 0003 4-monochlorobiphenyl 137273 -520 -519 -0018 24rsquo-dichlorobiphenyl 135764 -528 -600 07211 33rsquo-dichlorobiphenyl 135764 -580 -600 02012 34-dichlorobiphenyl 135764 -639 -600 -03915 44rsquo-dichlorobiphenyl 135764 -656 -600 -05618 22rsquo5-trichlorobiphenyl 134963 -602 -643 04131 24rsquo5-trichlorobiphenyl 134963 -625 -643 01837 344rsquo-trichlorobiphenyl 134963 -706 -643 -06344 22rsquo35rsquo-tetrachlorobiphenyl 133260 -647 -735 08852 22rsquo55rsquo-tetrachlorobiphenyl 133260 -700 -735 03566 23rsquo44rsquo-tetrachlorobiphenyl 133260 -668 -735 06775 244rsquo6-tetrachlorobiphenyl 133260 -694 -735 04177 33rsquo44rsquo-tetrachlorobiphenyl 133260 -853 -735 -11880 33rsquo55rsquo-tetrachlorobiphenyl 133260 -854 -735 -11983 22rsquo33rsquo5-pentachlorobiphenyl 133252 -696 -735 03986 22rsquo345-pentachlorobiphenyl 133252 -721 -735 01487 22rsquo345rsquo-pentachlorobiphenyl 133252 -791 -735 -05688 22rsquo346-pentachlorobiphenyl 133252 -743 -735 -008101 22rsquo455rsquo-pentachlorobiphenyl 133252 -733 -735 002104 22rsquo466rsquo-pentachlorobiphenyl 133252 -732 -735 003118 23rsquo44rsquo5-pentachlorobiphenyl 133252 -739 -735 -004128 22rsquo33rsquo44rsquo-hexachlorobiphenyl 131692 -901 -819 -082129 22rsquo33rsquo45-hexachlorobiphenyl 131692 -807 -819 012138 22rsquo344rsquo5rsquo-hexachlorobiphenyl 131692 -832 -819 -013151 22rsquo355rsquo6-hexachlorobiphenyl 131692 -742 -819 077183 22rsquo344rsquo5rsquo6-heptachlorobiphenyl 131261 -792 -843 051187 22rsquo34rsquo55rsquo6-heptachlorobiphenyl 131261 -894 -843 -051202 22rsquo33rsquo55rsquo66rsquo-octachlorobiphenyl 129924 -915 -915 000208 22rsquo33rsquo455rsquo66rsquo-nonachlorobiphenyl 127599 -1041 -1040 -001209 decachlorobiphenyl 125352 -1162 -1161 -001

Average absolute deviation = 038

Table 10a Modeling of the PCBs solubility with Eq (7) based on D(ao0EC) - training set

UnauthenticatedDownload Date | 82319 947 PM

EA Castro et al Central European Journal of Chemistry 2(3) 2004 500ndash523 519

n Molecule D(ao0EC) log Sw Eq (7) Residue

1 2-monochlorobiphenyl 137273 -454 -519 0654 22rsquo-dichlorobiphenyl 135764 -527 -600 07310 26-dichlorobiphenyl 135764 -521 -600 07922 234rsquo-trichlorobiphenyl 134963 -626 -643 01724 236-trichlorobiphenyl 134963 -629 -643 01426 23rsquo5-trichlorobiphenyl 134963 -601 -643 04228 244rsquo-trichlorobiphenyl 134963 -621 -643 02229 245-trichlorobiphenyl 134963 -627 -643 01630 246-trichlorobiphenyl 134963 -614 -643 02933 2rsquo34-trichlorobiphenyl 134963 -629 -643 01440 22rsquo33rsquo-tetrachlorobiphenyl 133260 -728 -735 00747 22rsquo44rsquo-tetrachlorobiphenyl 133260 -651 -735 08449 22rsquo45rsquo-tetrachlorobiphenyl 133260 -657 -735 07853 22rsquo56rsquo-tetrachlorobiphenyl 133260 -708 -735 02754 22rsquo66rsquo-tetrachlorobiphenyl 133260 -721 -735 01461 2345-tetrachlorobiphenyl 133260 -716 -735 01970 23rsquo4rsquo5-tetrachlorobiphenyl 133260 -725 -735 01082 22rsquo33rsquo4-pentachlorobiphenyl 133252 -705 -735 030116 23456-pentachlorobiphenyl 133252 -792 -735 -057134 22rsquo33rsquo56-hexachlorobiphenyl 131692 -860 -819 -041136 22rsquo33rsquo66rsquo-hexachlorobiphenyl 131692 -865 -819 -046141 22rsquo3455rsquo-hexachlorobiphenyl 131692 -768 -819 051153 22rsquo44rsquo55rsquo-hexachlorobiphenyl 131692 -856 -819 -037155 22rsquo44rsquo66rsquo-hexachlorobiphenyl 131692 -871 -819 -052156 233rsquo44rsquo5-hexachlorobiphenyl 131692 -782 -819 037158 233rsquo44rsquo6-hexachlorobiphenyl 131692 -766 -819 053171 22rsquo33rsquo44rsquo6-heptachlorobiphenyl 131261 -830 -843 013185 22rsquo3455rsquo6-heptachlorobiphenyl 131261 -846 -843 -003194 22rsquo33rsquo44rsquo55rsquo-octachlorobiphenyl 129924 -916 -915 -001206 22rsquo33rsquo44rsquo55rsquo6-nonachlorobiphenyl 127599 -1026 -1040 014

Average absolute deviation = 035

Table 10b Modeling of the PCBs solubility with Eq (7) based on D(ao0EC) - test set

UnauthenticatedDownload Date | 82319 947 PM

520 EA Castro et al Central European Journal of Chemistry 2(3) 2004 500ndash523

n Molecule D(ao1EC) log Sw Eq(8) Residue

- biphenyl 210491 -431 -430 -0013 4-monochlorobiphenyl 207973 -520 -519 -0018 24rsquo-dichlorobiphenyl 206194 -528 -582 05411 33rsquo-dichlorobiphenyl 206185 -580 -583 00312 34-dichlorobiphenyl 205306 -639 -614 -02515 44rsquo-dichlorobiphenyl 205075 -656 -622 -03418 22rsquo5-trichlorobiphenyl 205681 -602 -601 -00131 24rsquo5-trichlorobiphenyl 204562 -625 -640 01537 344rsquo-trichlorobiphenyl 203119 -706 -691 -01544 22rsquo35rsquo-tetrachlorobiphenyl 202897 -647 -699 05252 22rsquo55rsquo-tetrachlorobiphenyl 203245 -700 -687 -01366 23rsquo44rsquo-tetrachlorobiphenyl 201085 -668 -764 09675 244rsquo6-tetrachlorobiphenyl 201889 -694 -735 04177 33rsquo44rsquo-tetrachlorobiphenyl 200359 -853 -789 -06480 33rsquo55rsquo-tetrachlorobiphenyl 201793 -854 -738 -1168 22rsquo33rsquo5-pentachlorobiphenyl 201662 -696 -743 04786 22rsquo345-pentachlorobiphenyl 202421 -721 -716 -00587 22rsquo345rsquo-pentachlorobiphenyl 202019 -791 -730 -06188 22rsquo346-pentachlorobiphenyl 202499 -743 -713 -030101 22rsquo455rsquo-pentachlorobiphenyl 201479 -733 -750 017104 22rsquo466rsquo-pentachlorobiphenyl 202925 -732 -698 -034118 23rsquo44rsquo5-pentachlorobiphenyl 200036 -739 -801 062128 22rsquo33rsquo44rsquo-hexachlorobiphenyl 199268 -901 -828 -073129 22rsquo33rsquo45-hexachlorobiphenyl 199823 -807 -808 001138 22rsquo344rsquo5rsquo-hexachlorobiphenyl 198728 -832 -847 015151 22rsquo355rsquo6-hexachlorobiphenyl 200054 -742 -800 058183 22rsquo344rsquo5rsquo6-heptachlorobiphenyl 198958 -792 -839 047187 22rsquo34rsquo55rsquo6-heptachlorobiphenyl 198763 -894 -846 -048202 22rsquo33rsquo55rsquo66rsquo-octachlorobiphenyl 196833 -915 -914 -001208 22rsquo33rsquo455rsquo66rsquo-nonachlorobiphenyl 193267 -1041 -1041 000209 decachlorobiphenyl 189910 -1162 -1160 -002

Average absolute deviation = 033

Table 11a Modeling of the PCBs solubility with Eq (8) based on D(ao1EC) - training set

UnauthenticatedDownload Date | 82319 947 PM

EA Castro et al Central European Journal of Chemistry 2(3) 2004 500ndash523 521

n Molecule D(ao1EC) log Sw Eq(8) Residue

1 2-monochlorobiphenyl 209092 -454 -480 0264 22rsquo-dichlorobiphenyl 207313 -527 -543 01610 26-dichlorobiphenyl 207391 -521 -540 01922 234rsquo-trichlorobiphenyl 204214 -626 -653 02724 236-trichlorobiphenyl 205411 -629 -610 -01926 23rsquo5-trichlorobiphenyl 205117 -601 -621 02028 244rsquo-trichlorobiphenyl 203845 -621 -666 04529 245-trichlorobiphenyl 204076 -627 -657 03030 246-trichlorobiphenyl 204880 -614 -629 01533 2rsquo34-trichlorobiphenyl 204238 -629 -652 02340 22rsquo33rsquo-tetrachlorobiphenyl 202549 -728 -712 -01647 22rsquo44rsquo-tetrachlorobiphenyl 200359 -651 -789 13849 22rsquo45rsquo-tetrachlorobiphenyl 202528 -657 -712 05553 22rsquo56rsquo-tetrachlorobiphenyl 203887 -708 -664 -04454 22rsquo66rsquo-tetrachlorobiphenyl 204529 -721 -641 -08061 2345-tetrachlorobiphenyl 202582 -716 -710 -00670 23rsquo4rsquo5-tetrachlorobiphenyl 201802 -725 -738 01382 22rsquo33rsquo4-pentachlorobiphenyl 201671 -705 -743 038116 23456-pentachlorobiphenyl 202877 -792 -700 -092134 22rsquo33rsquo56-hexachlorobiphenyl 199706 -860 -812 -048136 22rsquo33rsquo66rsquo-hexachlorobiphenyl 200858 -865 -772 -093141 22rsquo3455rsquo-hexachlorobiphenyl 200171 -768 -796 028153 22rsquo44rsquo55rsquo-hexachlorobiphenyl 198188 -856 -866 010155 22rsquo44rsquo66rsquo-hexachlorobiphenyl 199796 -871 -809 -062156 233rsquo44rsquo5-hexachlorobiphenyl 198728 -782 -847 065158 233rsquo44rsquo6-hexachlorobiphenyl 198806 -766 -844 078171 22rsquo33rsquo44rsquo6-heptachlorobiphenyl 199498 -830 -820 -010185 22rsquo3455rsquo6-heptachlorobiphenyl 200941 -846 -769 -077194 22rsquo33rsquo44rsquo55rsquo-octachlorobiphenyl 197067 -916 -906 -010206 22rsquo33rsquo44rsquo55rsquo6-nonachlorobiphenyl 193384 -1026 -1037 011

Average absolute deviation = 040

Table 11b Modeling of the PCBs solubility with Eq (8) based on D(ao1EC) - test set

UnauthenticatedDownload Date | 82319 947 PM

522 EA Castro et al Central European Journal of Chemistry 2(3) 2004 500ndash523

n Molecule D(aNNC) log Sw Eq (9) Residue

- biphenyl 56206 -431 -433 0023 4-monochlorobiphenyl 55669 -520 -523 0038 24rsquo-dichlorobiphenyl 55179 -528 -604 07611 33rsquo-dichlorobiphenyl 55179 -580 -604 02412 34-dichlorobiphenyl 55179 -639 -604 -03515 44rsquo-dichlorobiphenyl 55179 -656 -604 -05218 22rsquo5-trichlorobiphenyl 54920 -602 -647 04531 24rsquo5-trichlorobiphenyl 54920 -625 -647 02237 344rsquo-trichlorobiphenyl 54920 -706 -647 -05944 22rsquo35rsquo-tetrachlorobiphenyl 54370 -647 -738 09152 22rsquo55rsquo-tetrachlorobiphenyl 54370 -700 -738 03866 23rsquo44rsquo-tetrachlorobiphenyl 54370 -668 -738 07075 244rsquo6-tetrachlorobiphenyl 54370 -694 -738 04477 33rsquo44rsquo-tetrachlorobiphenyl 54370 -853 -738 -11580 33rsquo55rsquo-tetrachlorobiphenyl 54370 -854 -738 -11683 22rsquo33rsquo5-pentachlorobiphenyl 54359 -696 -740 04486 22rsquo345-pentachlorobiphenyl 54359 -721 -740 01987 22rsquo345rsquo-pentachlorobiphenyl 54359 -791 -740 -05188 22rsquo346-pentachlorobiphenyl 54359 -743 -740 -003101 22rsquo455rsquo-pentachlorobiphenyl 54359 -733 -740 007104 22rsquo466rsquo-pentachlorobiphenyl 54359 -732 -740 008118 23rsquo44rsquo5-pentachlorobiphenyl 54359 -739 -740 001128 22rsquo33rsquo44rsquo-hexachlorobiphenyl 53860 -901 -823 -078129 22rsquo33rsquo45-hexachlorobiphenyl 53860 -807 -823 016138 22rsquo344rsquo5rsquo-hexachlorobiphenyl 53860 -832 -823 -009151 22rsquo355rsquo6-hexachlorobiphenyl 53860 -742 -823 081183 22rsquo344rsquo5rsquo6-heptachlorobiphenyl 53719 -792 -846 054187 22rsquo34rsquo55rsquo6-heptachlorobiphenyl 53719 -894 -846 -048202 22rsquo33rsquo55rsquo66rsquo-octachlorobiphenyl 53289 -915 -917 002208 22rsquo33rsquo455rsquo66rsquo-nonachlorobiphenyl 52524 -1041 -1044 003209 decachlorobiphenyl 51801 -1162 -1164 002

Average absolute deviation = 039

Table 12a Model of the PCBs solubility with Eq (9) based on D(aNNC) - training set

UnauthenticatedDownload Date | 82319 947 PM

EA Castro et al Central European Journal of Chemistry 2(3) 2004 500ndash523 523

n Molecule D(aNNC) log Sw Eq (9) Residue

1 2-monochlorobiphenyl 55669 -454 -523 0694 22rsquo-dichlorobiphenyl 55179 -527 -604 07710 26-dichlorobiphenyl 55179 -521 -604 08322 234rsquo-trichlorobiphenyl 54920 -626 -647 02124 236-trichlorobiphenyl 54920 -629 -647 01826 23rsquo5-trichlorobiphenyl 54920 -601 -647 04628 244rsquo-trichlorobiphenyl 54920 -621 -647 02629 245-trichlorobiphenyl 54920 -627 -647 02030 246-trichlorobiphenyl 54920 -614 -647 03333 2rsquo34-trichlorobiphenyl 54920 -629 -647 01840 22rsquo33rsquo-tetrachlorobiphenyl 54370 -728 -738 01047 22rsquo44rsquo-tetrachlorobiphenyl 54370 -651 -738 08749 22rsquo45rsquo-tetrachlorobiphenyl 54370 -657 -738 08153 22rsquo56rsquo-tetrachlorobiphenyl 54370 -708 -738 03054 22rsquo66rsquo-tetrachlorobiphenyl 54370 -721 -738 01761 2345-tetrachlorobiphenyl 54370 -716 -738 02270 23rsquo4rsquo5-tetrachlorobiphenyl 54370 -725 -738 01382 22rsquo33rsquo4-pentachlorobiphenyl 54359 -705 -740 035116 23456-pentachlorobiphenyl 54359 -792 -740 -052134 22rsquo33rsquo56-hexachlorobiphenyl 53860 -860 -823 -037136 22rsquo33rsquo66rsquo-hexachlorobiphenyl 53860 -865 -823 -042141 22rsquo3455rsquo-hexachlorobiphenyl 53860 -768 -823 055153 22rsquo44rsquo55rsquo-hexachlorobiphenyl 53860 -856 -823 -033155 22rsquo44rsquo66rsquo-hexachlorobiphenyl 53860 -871 -823 -048156 233rsquo44rsquo5-hexachlorobiphenyl 53860 -782 -823 041158 233rsquo44rsquo6-hexachlorobiphenyl 53860 -766 -823 057171 22rsquo33rsquo44rsquo6-heptachlorobiphenyl 53719 -830 -846 016185 22rsquo3455rsquo6-heptachlorobiphenyl 53719 -846 -846 000194 22rsquo33rsquo44rsquo55rsquo-octachlorobiphenyl 53289 -916 -917 001206 22rsquo33rsquo44rsquo55rsquo6-nonachlorobiphenyl 52524 -1026 -1044 018

Average absolute deviation = 037

Table 12b Model of the PCBs solubility with Eq (9) based on D(aNNC) - test set

UnauthenticatedDownload Date | 82319 947 PM