2-hydroxy-1-(4-(4-(2-hydroxy-2-methylpropionyl)benzyl)phenyl)-2-methylpropan-1-one

Henry's Law constant

Link to relevant study record(s)

Reference

Endpoint:

Henry's law constant

Type of information:

(Q)SAR

Remarks:

Migrated phrase: estimated by calculation

Adequacy of study:

key study

Study period:

5 January 2012

Reliability:

2 (reliable with restrictions)

Rationale for reliability incl. deficiencies:

accepted calculation method

Justification for type of information:

1. SOFTWARE
EPIWIN

2. MODEL (incl. version number)
version 3.2

3. SMILES OR OTHER IDENTIFIERS USED AS INPUT FOR THE MODEL
CC(C)(O)C(=O)c2ccc(Cc1ccc(C(=O)C(C)(C)O)cc1)cc2

4. SCIENTIFIC VALIDITY OF THE (Q)SAR MODEL

Bond Contribution Methodology
HENRYWIN estimates the Henry's Law Constant of organic compounds at 25oC using the methodology originally described by Hine and Mookerjee (1975).  The original bond  contribution methodology was updated and expanded as described in Meylan and Howard (1991).

The following abstract from the Meylan and Howard (1991) article briefly summarizes the bond methodology:  

"Bond contribution values, used to estimate Henry's law constant (HLC)(air-to-water partition coefficient) from chemical structure, have been determined for 59 chemical bonds by a least-square analysis of HLCs for 345 organic compounds.  A correlation coefficient (r2) of 0.94 was determined for the relationship between known LWAPCs (log water-to-air partition coefficients) and bond estimated LWAPCs for the 345 compound data set.  The correlation increases to 0.97 when quantified correction factors are applied to selected chemical classes.  The ability of the bond method to estimate LWAPCs is demonstrated by a validation test set of 74 diverse and structurally complex compounds that were not included in the least-square analysis.  The correlation coefficient for the validation set is 0.96."  (Note: the bond value and correction factor list has been expanded significantly since the journal article was published).

-Introduction:
The Henry's Law constant (HLC) can be defined as the ratio of the concentration of a compound in the gas-phase to the concentration of the compound in a dilute aqueous solution at equilibrium.  This ratio or partition coefficient is usually known as the unitless HLC.  Most literature reports experimental HLCs as either the unitless value or as a value with units of atm-m3/mole.  The unitless value is converted to units of atm-m3/mole by multiplying it by the gas constant (8.206 X 10-5 atm-m3/mole K) and the temperature (in deg K). Hine and Mookerjee (1975) reported HLC values as the logarithm of the reciprocal unitless value,  and HENRYWIN uses the same value for consistency.  However, to avoid confusion between HLC and the log reciprocal value, the latter is designated as LWAPC (Log Water-to-Air Partition Coefficient).

The methodology used to derive the bond contribution values is identical to that of Hine and Mookerjee (1991).  Each compound is split into a summation of the individual bonds which comprise the compound.  The summation is set equal to the compound's known LWAPC.  For example, ethanol is comprised of one C-C bond, five C-H bonds, one C-O bond, and one O-H bond.  Each different bond is assigned to a unique variable in a linear equation.  For ethanol, the linear equation is set equal to 3.690, the known LWAPC for ethanol at 25oC.  All compounds used to derive the bond contribution values are treated in a similar manner.  Selected functional groups, including the cyano (CN), carbonyl (CO), and nitro groups (NO2), are treated as atoms and are not split into individual bonds.  The known LWAPCs used in the equations are experimentally measured LWAPCs.  When a measured LWAPC is not available, the known LWAPC is derived from the measured VP-WS ratio.

All linear equations were subjected to a least-squares treatment (a multiple-linear regression) to determine the best fit for each bond contribution value.  The initial data set for the original MS-DOS version of HENRYWIN included 345 equations and 59 bond contribution values.

-Data Collection:
Experimental LWAPCs, vapor pressures, and water solubilities were obtained from Syracuse Research Corporation's Environmental Fate Data Base (EFDB) system (Howard et al, 1982, 1986) and associated literature surveys, including the data compilation of Hine and Mookerjee (1975).  Values at 25oC were used when possible.  A total of 427 chemicals were located for inclusion in the original study (Meylan and Howard, 1991).  From this total, a set of 345 chemicals was selected for the least-square analysis.  The reasons for selecting the chemicals in the least-square set were as follows: (a) to include all available bond contribution values, (b) to include a diverse selection of chemicals to adequately represent the most commonly occurring bonds, and (c) to select individual chemicals that contain a limited number of different functional groups.  A validation set of 74 chemicals (chemicals not included in the least-square analysis) was selected to test the ability of the "least-square fitted" bond values to estimate accurate LWAPCs.  The validation set includes many compounds that are structurally more complex than included the set used to derive the bond contribution values, thus providing an opportunity to evaluate the accuracy and rigor of the bond contribution method.

When possible, retrieved LWAPCs, water solubilities, and vapor pressures were critically evaluated for their reliability by examining the experimental technique used in their determination and by comparing different experimental values for the same compound from different sources.  Several data bases that already included evaluation codes were used to obtain various vapor pressures and water solubilities.  A substantial number of vapor pressures were obtained from the Daubert and Danner Data Compilation (1989) that includes a quality code of experimental accuracy.  Likewise, some water solubilities were taken from the ARIZONA DataBase of Aqueous Solubilities (Yalkowsky, 1989) that includes an evaluation code.  More than one experimentally measured LWAPC was available for approximately 20% of the compounds that had experimentally measured LWAPCs.  In most cases, the different measurements were relatively close.  For the compounds with close measurements, either a single source value or a median value was selected.  If the measurements are close, it does not make a significant difference which value is selected for the least-square analysis.  For the infrequent cases when significant differences existed between measured LWAPCs, the value from the most highly evaluated source or most reliable experimental method was selected.  Reliably evaluated measurements at 25oC were selected over measurements at other temperatures.  All but two selected vapor pressures are at 25oC.  One selected vapor pressure is at 30oC and the other is at 20oC; corresponding water solubilities at 30o and 20oC were used to calculate VP-WS LWAPCs.  Several water solubilities were selected at temperatures close to 25oC (e.g., 17-23oC) with corresponding vapor pressures at 25oC.  VP-WS LWAPCs for these selections were calculated with water solubilities corrected to 25oC, although, these water solubilities require little or no correction.

All vapor pressure, water solubility, LWAPC values, and their sources from the 345 chemical training set and 74 compound validation set are available via Internet download at:

http://esc.syrres.com/interkow/EpiSuiteData.htm

Appendix G lists the 345 compounds in the original regression with corresponding known and HENRYWIN estimated LWAPC values.

Appendix I lists the 74 compounds in the validation set.

Full reference citations are from the Experimental References section and the on-line PhysProp Reference help file.

-LWAPC Values from Vapor Pressure / Water Solubility:

The validity of estimating LWAPCs from the VP-WS ratio is clearly demonstrated by selected  values in the 345 chemical data set.  The 345 chemical data set contains a subset of 120 compounds that have both an experimentally measured LWAPC and measured vapor pressures and water solubilities.  Compounds that are completely miscible in water are not included in the subset.  The correlation coefficient (r2) for VP-WS LWAPC to experimental LWAPC for this subset is 0.997 with a absolute mean error and standard deviation of 0.05 and 0.07 log units, respectively.  Clearly, the VP/WS ratio is an excellent estimation method when reliable data are available.

It has been recommended that estimating HLCs from the VP-WS ratio be limited to compounds with water solubilities less than 1 mole/liter (Lyman, 1985).  Data from the 345 chemical data set indicate that most compounds with water solubilities as high as 10-30 % (g/100g) are adequately estimated by the VP-WS ratio.

-Bond Contribution Multiple-Linear Regression:
The 345 compound training set was placed in a matrix of 345 rows by 60 columns ...  59 of the 60 columns corresponded to each of the 59 bond contributions containing the number of instances of bond contribution for each compound.  The 60th column was the solution column containing the known LWAPC for each compound.  The "best-fit" solution of the matrix was accomplished by least-square analysis.  The original analysis was done with Fortran computer code (Press et al., 1985) and then repeated with commercial statistical software (CoHort, 1988).

Appendix D lists the bond contribution values derived by the multiple-linear regression. Appendix D also lists the bond contribution values derived in a subsequent update of the HENRYWIN program.  In addition, a variety of "Estimated" bond contribution values have been added to HENRYWIN based on limited experimental data or analogy to similar bonds (see Appendix D for a list and additional information).

-Correction Factors:
Various chemical classes do not correlate as well as others in the bond contribution method.  Many deviations observed in the bond contribution correlation can be attributed to "polar interactions" (Hine and Mookerjee, 1975).  Although Hine and Mookerjee discussed the magnitude of various "polar interactions", they did not attempt to apply any correction factors to attenuate the deviations.  The Meylan and Howard (1991) study did derive correction factors used in HENRYWIN.

The correction factors are simply mean values of the deviations between known and estimated LWACPs.  When appicable, a correction factor is added to the summed total of the bond contribution values of a compound.  The correlation coefficient (r2) increases to 0.97 (from 0.94) and the mean error and standard deviation fall to 0.21 and 0.34 log units (from 0.30 and 0.45), respectively, when correction factors are applied to the 345 chemical data set.  Eighty-one of the 345 chemicals had applicable correction factors.

Appendix E lists the bond method correction factors determined in the Meylan and Howard (1991) study.  It also lists the correction factors (determined via multiple-linear regression of 90 compounds) in a subsequent update of HENRYWIN.

-Bond Contribution Method SAR Equation:
The final bond contribution method SAR equation for HENRYWIN is:

LWAPC = Σ((Bi)(Nj) + (Ci)(Mj))

LWAPC is the summation of the bond contribution value of each bond (Bi)  times the number of instances of each bond (Nj) plus the correction factor value of each factor (Ci) times the number of instances of each correction factor (Mj).

-Estimation Accuracy:
See the Accuracy section

-Esimation Domain:
Appendix D and Appendix E give for each bond contribution and correction factor used in regressing the Bond Methodology the maximum number of instances of that bond or correction factor in any of the training set compounds (the minimum number of instances is of course zero, since not all compounds had every fragment).  Currently there is no universally accepted definition of model domain.  However, users may wish to consider the possibility that estimates are less accurate for compounds outside the MW range of the training set compounds, and/or that have more instances of a given bond or correction factor than the maximum for all training set compounds.  It is also possible that a compound may have a functional group(s) or other structural features not represented in the training set, and for which no bond or correction factor coefficient was developed.  These points should be taken into consideration when interpreting model results.

Both the Bond and Group contribution methods of predicting Henry's law constants can yield estimates resulting in "Missing Fragments" (an example is shown in the Missing Fragments section of the Group Methodology chapter).  When this occurs, the HENRYWIN estimate is labeled "Incomplete".

Ranges for the 442 Compound Dataset (Appendix G) used for Regressing the Bond Method coefficients (via least-square analysis):

- Molecular Weight:

  Minimum:  26.04 (Ethyne)

  Maximum:  451.47 (Flucythrinate, 70124-77-5)

  Average:  144.64

- Henry's law constant (atm-m3/mole):

  Minimum:  5.65x10-14 (Karbutilate (4849-32-5); LWAPC = 11.636)

  Maximum: 2.03x10+1  (Hexafluoroethane; LWAPC = -2.919)

Principles of method if other than guideline:

Calculation using SRC HENRYWIN v3.20
Bond estimation

GLP compliance:

no

H:

0.012 Pa m³/mol

Temp.:

25 °C

Conclusions:

H = 0.0122 Pa.m^3.mol^-1 at 25 °C (Bond estimation method)

Executive summary:

In this validated QSAR study (SRC HENRYWIN v3.20) the test material (EC 444-860-9) was determined to have a Henry's Law constant of H = 0.0122 Pa.m³.mol^-1 at 25°C using the Bond estimation method.

Description of key information

Study conducted using validated QSAR model with suitable applicability domain.

Key value for chemical safety assessment

Henry's law constant (H) (in Pa m³/mol):

0.012

at the temperature of:

25 °C

Additional information

The test substance will not significantly evaporate into the atmosphere from the water surface. Henry´s law constant (25°C) was estimated to be 0.0122 Pa m³/mol by SRC HENRYWIN v3.20 [BASF 2012].