Fifth Simulation Challenge

Prediction of 1-Octanol-Water Partition and Infinite-Dilution Activity Coefficients

The objective of this Challenge of the IFPSC is to test the ability of computer modeling (any method) to predict 1-Octanol-Water Partition (ko/w) and Infinite-Dilution Activity Coefficients (gam-inf). Prediction of 1-Octanol-Water Partition Coefficients (which, admittedly, has been widely studied by non-molecular simulation computational methods) is viewed as a stepping stone to a more difficult problem, such as a case where a third component is present at a high enough level to influence the mutual solubilities of the other two.
April 7, 2008 - Problem announced
September 30, 2008 - Entries due; to be submitted to 
November 2008 - Champions announced
The ability of computer modeling to predict properties that are challenging, inaccessible to experiment, or simply missing is often used as a justification for its development. We want to test/promote/validate this capability via the 5th Simulation Challenge. The 5th Simulation Challenge is similar in spirit to a growing number of modeling and simulation challenges including for example the recent QSAR World modeling challenge for human oral absorption (
Thermophysical properties, phase equilibria, and solution chemistries are the underlying physical and chemical phenomena of industrial chemical processes. Rigorous thermodynamic modeling of such phenomena establishes a sound and scientific foundation for simulation of chemical processes and subsequent process development, optimization and control.
There are three main aspects of applied thermodynamics and phase equilibria of interest:
  1. How to obtain data experimentally?
  2. How to predict properties of pure components or mixtures of components?
  3. How to correlate limited data so they can be interpolated or extrapolated or combined into a representation of multicomponent behavior?
In the first 4 challenges we have tested these aspects with varying results. For example, one task was to predict pure component vapor pressures. Another was to predict vapor-liquid equilibria (VLE) of binary mixtures at varying conditions.
In this 5th challenge we want to refocus on aspect number 2. It is a common task in chemical and related industries to use octanol-water partition coefficients to aid in the design of product formulations with desired hydrophobicity or hydrophilicity character. ko/w is also used to estimate environmental bio-accumulation factors because it provides a crude model of partitioning between blood (water) and body tissue (1-octanol). Infinite-dilution activity coefficients are used to calculate air-water partition coefficients (Henry's Law constants) for environmental engineering applications. In addition, VLE of binary mixtures can be predicted (interpolated) over the entire binary composition using a two-parameter activity-coefficient model if the two infinite-dilution activity coefficients are known (Prausnitz,, Molecular Thermodynamics of Fluid-Phase Equilibria, Third Edition, Appendix F, pps 804-807).
For each of the following molecules:
  • 1-ethylpropylamine, CAS# 616-24-0
  • 3-methyl-1-pentanol, CAS# 589-35-5 
Compute the:
  • 1-octanol-water partition coefficient (mole fraction units, assume neutral species) at 300 K and 101.325 kPa.
  • Infinite-dilution activity coefficient (mole fraction units, Lewis and Randall reference state) for the organic molecule dilute in water at 325 K and 13.5 kPa.
Rules of the game:
  • Any theory/modeling/simulation method, e.g. group contribution methods, can be used except very well established standard QSAR, group contribution, and related methods/tools. A list of such methods/tools and additional information can be found in the Appendix below.
  • The entries will be judged in two separate categories: molecular simulation methods and non-molecular simulation methods, respectively.
  • Any force field (or other model parameterization) previously published in the open literature prior to the announcement of this challenge is acceptable.
  • Force fields (or other models) that have not been published previously may not be parameterized for this challenge using octanol/water partition coefficient, infinite dilution activity coefficient or solubility in water data corresponding to the 2 challenge molecules. Force fields (or other models) may be parameterized using any other published physical property data.
  • Estimates of the uncertainty for computed water-octanol partition coefficients and the infinite-dilution activity coefficients must be included.
  • Participants may choose to include or exclude the effects of the known water saturation in 1-octanol. A recommended experimental saturation mole fraction of water in water-saturated (wet) 1-octanol is 0.27 from an experimental range of values of 0.20 to 0.29. (Sangster, J. Octanol-Water Partitioning Coefficients: Fundamentals and Physical Chemistry; John Wiley & Sons: Chichester, U.K. 1997)
Challenge Scoring:
Quantitative Accuracy Metrics (60%)
  • For each property, full credit will be awarded for predictions within the uncertainty limit of 5% of the experimental value.
  • A linear interpolation of partial credit will be awarded for predictions with an absolute deviation above the minimum threshold and a maximum of 40%. No points will be awarded for prediction above the maximum deviation.
Relative Accuracy Metrics - Relative Ranking (40%):
The computed octanol-water partition coefficient and the infinite-dilution activity coefficient of each molecule will be compared to the respective experimental values. The molecule that is in best agreement to experiment will be used as a normalization to determine relative coefficients (ie: (coefficient worst predicted)/(coefficient best)). The relative ratios will be compared to the corresponding normalized ratios using the experimental data. The quantity (coefficient/coefficient-ref) predicted will receive full credit if it is within 5% of the corresponding quantity (coefficient/coefficient-ref)expt. As in the quantitative accuracy section, similar partial credit will be awarded on a sliding scale.
Click on the link below to see the 5th Challenge Problem scoring sheet.
Other entry guidelines:
  • Entries are to be submitted to on or before the deadline
  • A submission for this challenge problem is to be in the form of a manuscript suitable for submission to a refereed, archival, scientific journal. The manuscript must contain sufficient detail about the simulation or other method and about the force field (if simulation) so that an experienced simulator could reproduce the results without requiring access to proprietary information. In particular, all potential parameters and molecule geometry parameters must be explicitly specified in the manuscript. The results are to be reported in SI units.
  • An analysis of the uncertainty in the calculated results is required and must be included in the manuscript.
  • Entries are expected to present results that are statistically significant and to present sufficient supporting evidence to establish this quality. Also, the scientific reasoning behind any new (unpublished) force field parameterizations must be clearly spelled out in the entry. If there is a consensus among the judges that an entry is of poor quality (uses a method commonly accepted to be fundamentally flawed, presents results that are not statistically significant, fails to provide sufficient supporting data and details, violates the various rules and guidelines established for the competition, or for any other reason would be unlikely to be accepted by any peer-reviewed scientific journal in the field), that entry will be rejected and will not be considered in the judging.
  • Entries that represent collaborations between multiple research groups are welcomed.
Standard QSAR (and related methods) and Group Contribution Estimation Methods for Octanol-Water Partition Coefficients and Infinite-Dilution Activity Coefficients
Use of standard QSAR (and related methods) and group contribution methods is not allowed in the Fifth Challenge. The Fifth Challenge Benchmark Data Committee will provide values from these standard methods for comparison.
The standard octanol-water partition coefficient QSAR (and related) methods include those shown on the Virtual Computational Chemistry Laboratory web-page: . This page contains links to seven standard QSAR methods that are not allowed for use in the Fifth Challenge. Here are the web-sites for the seven disallowed methods:
  1. ALOGPs
  2. AC LogP
  3. AB/LogP
  4. COSMOFrag
  5. MiLogP
  6. KOWWIN 
  7. or
  8. XLogP
Other published and/or commercially available standard QSAR methods, such as CLogP (, are also not allowed.
Entries based on a new QSAR method that is developed in response to the Fifth Challenge will be allowed. See J. Phys. Chem. Ref. Data 36, 2007, 203 for an example of development of a specialized QSAR method.
The use of published group-contribution-based activity coefficient estimation methods as a route to octanol-water partition coefficients or infinite-dilution activity coefficients is also disallowed from the Fifth Challenge. These methods include the original and published variations of UNIFAC (AICHEJ 21, 1975, 1086) and ASOG (Ind. Eng. Chem. Fundam. 1, 1962, 20).
Entries based on a new group-contribution method that is developed in response to the Fifth Challenge will be allowed.
Q: Why did you exclude COSMOfrag as an established group contribution method. COSMOfrag is COSMO-RS based on sigma profiles which are on the fly taken from precalculated COSMO files? It has nothing to do with group contribution methods. 
A: The conditions of contest are designed to prohibit the use of automated, widely distributed KOW methods; this is: 1) to prevent a large number of identical entries, and 2) to promote original research that will demonstrate improved methodologies for KOW. COSMOFrag is one of several methods available on the web for automated calculation of a KOW value for any molecule submitted via SIMILES notation. We want to prohibit entries based on a 20-second calculation from this web-site; that is why COSMOFrag was listed as a prohibited method.
Q: Based on the results of the 4th challenge, why not focus on vapor phase viscosity (which would represent a major methodological advance)? 
A: While it does present some issues from a computation standpoint, it is not a property that had great industrial importance. When needed, other methods are available that give reasonably good estimates for vapor viscosity.
Q: Why not have a bigger scope by including a larger list of molecules covering a range of functionalities? 
A: While it would be more interesting if the problem incorporated a wide range of test molecules, a practical difficulty is the effort and cost of doing the measurements for a large number of molecules as opposed to a small number. Based on earlier feedback, a similar difficulty arises in the practical effort an entrant can devote to the problem. Overall, we're looking to balance scientific interest with practical industrial interest with what is a practical effort for an entrant to devote with what is practical to do in terms of new experimental measurements.
Q: Partitioning between octanol and water is complicated by the fact that water is soluble in octanol at a level of 0.2 to 0.3 in terms of mole fractions. Must entrants take this into account explicitly? Do they have to calculate the water saturation? Can they just ignore this complication? 
A: Entrants are not required to take into account water saturation of the octanol phase but may do so if they so desire. They are not required to calculate water saturation and can ignore this complication if they wish. This aspect illustrates a tension that exists in the IFPSC between the interests of science and practical industrial interests. A method that does not intrinsically produce the saturation composition for water/1-octanol will miss a key aspect of the science. On the other hand, a method that (given as input the saturation composition or treating neat octanol) could rapidly crank out accurate values of kow for a several different molecules would have great practical industrial value even if it is missing a big part of the science.
Q: The octanol-water partition coefficient has been widely studied by a variety of methodologies. Why not pick molecules from a different binary system or a different property?
A: The prediction of the 1-Octanol-Water Partition Coefficient is seen as a stepping stone to a more difficult problem, such as a case where a third component is present at a high enough level to influence the mutual solubilities of the other two. Methods to measure the 1-Octanol-Water Partition Coefficient are also well established and straight forward to carry out to obtain high quality experimental data for comparison to predicted values.
Q: There are good increment methods available as CLOGP. Hence participants would have the predicted CLOGP values as de-facto experimental values available and then could tune their methods until they achieve results close to those of CLOGP. Thus they would be quite sure that they are no more off than about 0.3 log units. Does the challenge expect higher accuracy than 0.3 log units.
A: 0.3 log units is a factor of 2 which is outside of the 40% limit to receive credit. The challenge is to predict the octanol-water partition coefficient to higher accuracy then this.
Subscribe to Comments for "Fifth Simulation Challenge"