Solubility FEP+ Best Practices

The solubility FEP+ workflow calculates the kinetic solubilities of amorphous solids and/or the thermodynamic solubilities of crystals. The solubility (desolution free energy) is calculated as the sum of the hydration free energy and the sublimation free energy. The following best practices should be deployed when using the solubility FEP+ workflow in projects.

  1. Solubility FEP+ can only calculate the solubility of solids (either amorphous or crystalline) containing the pure neutral ligand. If the solids contain counterions, cosolvents, or are cocrystals, we can not calculate the solubility of these solids with the currently available solubility FEP+ workflow.

  2. Before launching the solubility FEP+ workflow, perform Epik and/or Jaguar macropKa calculations to calculate the pKa of the molecules. For acidic or basic compounds with non-negligible population in charged form in aqueous solution, the pure neutral form in solid should be used as input for solubility FEP+, and the calculated solubility should be corrected according to the following formula taking into consideration the ionization equilibrium in aqueous solution: (S is the free energy to transfer from solid to solution, in kcal/mol)

    • for acidic compounds:

    • for basic compounds:

  3. Prepare the input structures for solubility FEP+ calculations.

    • For calculating the kinetic solubility of amorphous solids, prepare a low energy conformer for each ligand. ConfGen is recommended but other tools like MacroModel conformational sampling might also suffice.

    • For calculating the thermodynamic solubility of crystalline solids, prepare a supercell with ~100–200 molecules in a cubic shape. This can be done by replicating the unit cell of the crystals along the three principal axis. For most crystals of druglike compounds in the P21/c space group, a 3×3×3 supercell with 4×27=108 molecules suffice.  

  4. Use the Force Field Builder to check whether there are any torsions in the ligands not covered by the default OPLS4 parameters and refit them as needed.

  5. After the solubility FEP+ job finishes, inspect the Bennett errors of the five sublimation free energies and the hydration free energy.

    • If the Bennett error for hydration free energy is larger than 0.2, it may be advisable to increase the number of lambda windows and/or extend the hydration leg to improve convergence;
    • If the Bennett error for any of the five sublimation free energies is larger than 0.3, it is advisable to remove that sublimation free energy from the final calculations of the median sublimation free energy, rerun solubility FEP+ with more lambda windows or launch another independent solubility FEP+ calculation with a different random seed until obtaining at least 5 sublimation free energies with Bennett error less than 0.3.

  6. After confirming the Bennett errors of the hydration free energy and the sublimation free energies are in acceptable ranges, analyze the distribution of the five sublimation free energies. Ideally, they should be Gaussian distribututed and within 2–3 kcal/mol of each other. If the range is much wider than 2–3 kcal/mol and one or two sublimation free energies significantly deviate from others, extend the simulation times to obtain a better convergence, or launch additional independent sublimation free energy calculations and discard those that significantly deviate from others in the finial calculation of the median sublimation free energy;

  7. Enantiomers should have very similar hydration free energies and solubility. If there are enantiomers in the ligand series, but the hydration free energies differ by more than 0.3 kcal/mol between the two isomers or the solubility differs by more than 0.6 log units, it indicates convergence issues, and the simulations should be extended, or repeated with more lambda windows until reaching convergence.