Pure compound property prediction¶

Introduction¶

The Property Prediction program in ADF provides quick, accurate estimates for many important pure component physical properties. At its core, the Property Prediction program maps various QSPR descriptors of an input molecule onto a single numerical value, the property estimate. Many of these property models rely on easy-to-evaluate QSPR descriptors and numerically straightforward computations, meaning that an estimate can be provided for every property in << 1s per molecule. The general expression for the models used in the Property Prediction program is as follows:

where f is a function that transforms the property value space, g is a function that maps QSPR descriptors onto a numerical value, and h is a function which also captures temperature-dependence of certain properties by including temperature, T, as an input. Additionally, C is a constant, \(n_i\) refers to QSPR values of QSPR descriptor i, and \(c_i\) and \(d_i\) are fitted coefficients corresponding to each QSPR descriptor i.

The accuracy of the property estimates depends on the nature/complexity of the input molecular structure. For many common organic structures, the property estimates should be reasonably accurate. However, as is always the case with QSPR models, the Property Prediction program will likely lose accuracy for molecules outside its training domain, i.e., for molecules that are very “dissimilar” to compounds which occur in the training set. In general, the program can be used for molecules with the following atom types:

A brief description of molecule types for which this method will not work well is given in the General warnings section. Common molecules for which this method will fail are: (1) those that contain only one non-hydrogen atom, e.g., Methane or Water; (2) those that contain atoms not listed in the table above.

Available properties¶

The Property Prediction program can predict the values of various pure component physical properties. These properties can be of interest themselves or can be used in conjunction with other COSMO-RS property calculations (e.g., to calculate the solubility of a solid in a liquid, we must know the enthalpy of fusion and melting point of the solid – both of these properties can be estimated with the Property Prediction program). The available properties and their units are listed below:

Running the Property Prediction program¶

The Property Prediction program can be run from the command line. The following general flags are used by the program:

Note, if no output flag is supplied, then the results are written to a file called CRSKF by default. Additionally, the user may enter as many compounds as desired on the command line in either of the two available input formats.

The program can be run in 2 ways:

• Estimating all available properties for every molecule

• Estimating specific properties for every molecule

To estimate all properties for every input compound, simply execute the program with all molecules specified on the command line. Don’t forget that the -d flag is required to display the results in the terminal. An example of this is below.

$AMSBIN/prop_prediction --smiles CCCCCCO -o example.crskf -temperature 298.15 -temperature 398.15 -n 20 -d

Boiling point at standard pressure :
CCCCCCO              435.777  K
Critical pressure :
CCCCCCO              34.3493  bar
Critical temperature :
CCCCCCO              576.466  K
Critical volume  :
CCCCCCO             0.404124  L/mol
Liquid density :
CCCCCCO              0.79182  kg/L
Dielectric constant :
CCCCCCO              10.9512
Absolute entropy of an ideal gas at 298.15 K and 1 bar :
CCCCCCO              439.885  J/(mol K)
Flash point :
CCCCCCO              342.271  K
Gibbs energy of formation for an ideal gas at 298.15 K and 1 bar :
CCCCCCO             -131.869  kJ/mol
Net enthalpy of combustion at 298.15 K  :
CCCCCCO             -3678.12  kJ/mol
Standard state enthalpy of formation at 298.15 K and 1 bar :
CCCCCCO             -384.388  kJ/mol
Enthalpy of fusion at normal melting point :
CCCCCCO              18.5054  kJ/mol
Enthalpy of formation for an ideal gas 298.15 K :
CCCCCCO             -316.821  kJ/mol
Enthalpy of sublimation :
CCCCCCO              80.9799  kJ/mol
Melting point at 1 atm :
CCCCCCO              231.141  K
Liquid molar volume :
CCCCCCO             0.128949  L/mol
Parachor :
CCCCCCO              289.059
Solubility parameter :
CCCCCCO              10.1294  (cal/cm^3)^0.5
Triple point temperature :
CCCCCCO              230.404  K
Van der Waals area :
CCCCCCO              171.059  Å^2
Van der Waals volume :
CCCCCCO              120.519  Å^3
Liquid viscosity :
Molecule: CCCCCCO
   Temperature (K)     Liquid viscosity (Pa-s)
    298.15                 0.004465
    303.15                 0.003865
    308.15                 0.003364
    313.15                 0.002942
    318.15                 0.002584
    323.15                 0.002281
    328.15                 0.002021
    333.15                 0.001798
    338.15                 0.001606
    343.15                 0.001439
    348.15                 0.001294
    353.15                 0.001168
    358.15                 0.001057
    363.15                 0.000959
    368.15                 0.000873
    373.15                 0.000797
    378.15                 0.000729
    383.15                 0.000669
    388.15                 0.000615
    393.15                 0.000567
    398.15                 0.000523
Liquid vapor pressure :
Molecule: CCCCCCO
   Temperature (K)     Vapor pressure (bar)
    298.15                 0.001229
    303.15                 0.001809
    308.15                 0.002623
    313.15                 0.003750
    318.15                 0.005289
    323.15                 0.007362
    328.15                 0.010123
    333.15                 0.013757
    338.15                 0.018487
    343.15                 0.024582
    348.15                 0.032357
    353.15                 0.042182
    358.15                 0.054484
    363.15                 0.069757
    368.15                 0.088563
    373.15                 0.111537
    378.15                 0.139394
    383.15                 0.172929
    388.15                 0.213022
    393.15                 0.260644
    398.15                 0.316852

For most applications, it is not necessary to calculate all of the available physical properties (although doing so is practically just as fast). In these cases, additional property flags need to be specified on the command line to restrict the program to calculating only certain physical properties. For example, if we were doing solid/liquid solubility calculations on Ibuprofen and Paracetamol, we would require the Enthalpy of Fusion and the Melting Point of both compounds. To calculate only these two properties, we simply have to add the two property flags “-hfusion” and “-meltingpoint” to the command line. Using the .mol file for Ibuprofen and the SMILES string for Paracetamol, we execute the following:

$AMSBIN/prop_prediction -d -m Ibuprofen.mol -s 'CC(=O)NC1=CC=C(C=C1)O' -hfusion -meltingpoint

Enthalpy of fusion at normal melting point :
CC(=O)NC1=CC=C(C=C1)O      33.0298  kJ/mol
Ibuprofen.mol              24.0336  kJ/mol

Melting point at 1 atm :
CC(=O)NC1=CC=C(C=C1)O      469.282  K
Ibuprofen.mol              331.887  K

Index of property keys¶

The available properties and their corresponding property flags are listed below:

General warnings¶

This method will fail for the following types of molecules:

• Those that contain only one non-hydrogen atom (e.g., Methane or Water). However, experimental data is ample for these small molecules. The vapor pressure model is the exception in that it can represent such small structures.

• Those that contain atoms or substructures that are not listed in the Accepted atom types table above.

• Polymers and Ionic Liquids

This method will lose accuracy for some properties in the following domains:

• Molecules with many different types of functional groups

• Molecules that are extremely light (< 3 non-Hydrogen atoms) or heavy (> 30 non-Hydrogen atoms)

Equations for temperature-dependent properties¶

References¶

References

1

16. Ertl and A. Schuffenhauer. Estimation of synthetic accessibility score of drug-like molecules based on molecular complexity and fragment contributions Journal of cheminformatics 1 (2009): 1-11.