PerMM Instructions


To calculate the transmembrane translocation pathway and permeability coefficient, please define the temperature (in K) and pH and upload the coordinate file of a compound. The calculations take 1 to 4 minutes, depending on the number of conformatons included in the input coordinate file and the size of the molecule.

Two optimization options can be defined by user for calculation of the lowest energy pathway ∆G(z) of the molecule across the membrane:

  1. Drag method. The molecule is pulled across the membrane with local energy minimization with respect to rotational variables of the molecule in every point z +∆z, starting from the optimal rotational orientation found in the previous point z. This method produces an asymmetric energy curve and allows more precise estimation of the transmembrane energy barrier, especially for large amphiphilic molecules.
  2. Global energy optimization of rotational orientation of the molecule in each position z along the bilayer normal.

An option for ionizable compounds. The box at the bottom allows choosing an option ("yes") to assume an ionizable compund is uncharged in water. This option should be used to calculate the "intrinsic" BBB and CACO2 permeabilities (Po) or make comparisons with experimental BLM data provided for uncharged forms of ionizable molecules. Calculations with "no" option include energy of deionization of the molecule when it is transferred from water to nonpolar interior of the lipid bilayer. This energy depends on the difference of pH and pKa of ionizable groups. Note that "Calculated passive permeability" field in the PerMM database provides results of clculations for unionized state (i.e. Po).

Input coordinate files must be prepared in the PDB format. The coordinate files for peptides and small molecules can be downloaded from the Protein Data Bank, the Cambridge Structural Database, PubChem, DrugBank, or generated computationally using a variety of molecular modeling software (i.e. Chem3D). Converting SDF to PDB formats can be done by several on-line programs, including PyMOL open source visualization system (PyMOL). One can also use "Source" coordinate files provided by our PerMM database (they are downloadable from pages for individual molecules included in the database, or as a single .tar file from PerMM download page).

Multiple conformations of the molecules can be included as several models (MODEL 1, etc.; see "Source" file for Tol-Gly-Gly in PerMM database, for example). Lowest transfer energy conformer is selected automatically for vizualization in every point z of the curve, but the energy for calculating the permeability coefficient is averaged for the conformational ensemble.

Two slightly different versions of input PDB format are used for peptides and small molecules. The version for peptides uses standard PDB format. Hydrogen atoms can be included, but not required. Only all non-hydrogen atoms must be present in the file. The parameters of amino acid residues are taken from a library that includes all standard L- and D- residues, L-ornithine, D-Pen and a number of others. Any molecule with more than one residue indicated in standard PDB RESIDUE field will be automatically interpreted as a "peptide".

The coordinate files for small molecules must include all hydrogens, provide different names for all non-hydrogen atoms (e.g. N1,N2,N3...), and include pKa of all groups that are ionized within the pH values of interest (for example, 6.0 to 8.0). Small molecules must have the same residue number in RESIDUE field for all its atoms. A peptide can be defined as a small molecule by assigning the same residue number to all atoms, making sure that all non-hydrogen atoms have different names and by including the pKa values for ionizable molecules.
Using a PDB file without any pKa values will produce result for the unionized form. The pKa values must be assigned to non-hydrogen atoms of the corresponding ionizable group, such as O of COO- group or N of NH3+ group, and be included at the beginning of the file as REMARK PKA records (see example below and "Source" files of ionizable molecules included in PerMM database).

REMARK PKA O1 0 1.00 3.47 -1
ATOM 1 C ACETA 0 -0.116 0.604 0.376
ATOM 2 C1 ACETA 0 -0.806 -0.566 0.049
ATOM 3 C2 ACETA 0 -0.751 1.838 0.405
ATOM 4 C3 ACETA 0 -2.163 -0.646 -0.261
ATOM 5 C4 ACETA 0 -2.101 1.919 0.095
ATOM 6 C5 ACETA 0 -2.084 0.768 -0.238
ATOM 7 C6 ACETA 0 -0.137 -1.977 0.028
ATOM 8 C7 ACETA 0 2.137 0.727 -0.281
ATOM 9 C8 ACETA 0 3.583 0.610 0.202
ATOM 10 O ACETA 0 1.252 0.561 0.691

In this example, pKa value is assigned to one of oxygens in a COO- group; “O1” is the name of the atom to which the pKa value was assigned, “0” is the number of amino acid residue to which this atom belongs in the pdb file; “1.0” indicates that pKa was assigned to the single specified atom, rather than being distributed between several ionizable atoms (as a slightly better approximation, the user can include two O atoms of COO- group with weights 0.5 and 0.5, three nitrogens of guanidine group with weights of 0.33, etc.); “3.47” is the pKa value; and ”-1” is the sign of the charge at neutral pH (negatively charged group).

The remark for an NH3+ group would be, for example,

REMARK PKA N 0 1.00 3.47 -1

Importantly, the PDB format uses a position-dependent input, i.e. right-justified fields for numbers and left-justified fields for text. Hence the empty spaces in REMARK record are important and should be used as in the example above).

Dipole moments of polar groups are defined automatically, using a library of dipole moments for different chemical groups. Dipole moments are assigned to specific chemical groups, rather than to whole molecules. For example, a single dipole moment of Trp aromatic ring is automatically assigned to its N atom. A user can redefine group dipole moments manually (based on experimental or quantum chemistry-based values), which may improve the precision of calulations, especially for molecules with complex aromatic heterocyclic rings. Group dipole moments can be redefined by including the following records in the beginning of the PDB file (see example below or the source file for 2,3-Dideoxyadenosine in PerMM database). Note that the underlying PPM model was not parameterized for hydrophobic ions.

REMARK DIP N4 1 3.00
ATOM 1 N ADEN 1 -1.231 -1.450 0.000
ATOM 2 N1 ADEN 1 -2.074 0.637 0.000
ATOM 3 N2 ADEN 1 0.090 1.831 0.000
ATOM 4 N3 ADEN 1 1.992 0.396 0.000
ATOM 5 N4 ADEN 1 1.806 -1.912 0.000
ATOM 6 C ADEN 1 -0.162 -0.570 0.000

In this example, “N4” in the REMARK record is the name of the polar atom to which the dipole moment was assigned; “1” is the number of the “amino acid residue” where this atom belongs; and “3.00” is the group dipole moment (in D).


Output of the server includes the following:

  1. Graphical representation of the calculated transfer energy curve, ∆G (z)
  2. Logarithms of permeability coefficients are calculated for DOPC bilayer and estimated for plasma membrane, blood-brain barrier (BBB) and CACO2 cells.
  3. Interactive 3D visual images of a given compound along the translocation pathway ("Pathway" link) and the compound bound to the lipid bilayer ("Bound state" link). The dots in images (DUMMY atoms in the coordinate files) represent hydrocarbon core boundaries of the lipid bilayer; lipid headgroups are situated mostly outside these boundaries.
  4. Downloadable output coordinate file XXXout.pdb provides multiple positions of the molecule during its movement across the lipid bilayer. They are included as multiple models. Output coordinate file XXX1.pdb provides membrane-bound state of the molecule. Only the lowest transfer energy conformer in each position was included in the output file.
  5. Output messages including the calculated permeability coefficients and ∆G(z) curve. The curve also shows the number of the lowest energy conformer selected in every point of the curve.

Interpretation of Results

  1. The permeability coefficient is calculated for a monomeric molecule assumed to be completely soluble in water. The results are dependent on pH and temperature.
  2. Permeability coefficients LogP are calculated for BLM (DOPC bilayer) and roughly estimated for plasma membrane (PM), blood-brain barrier (BBB) and CACO2 cells. A significant deviation of values for BBB and CACO2 from values for PM indicate the influence of faclitated diffusion, transport, efflux or adsorption in vivo. Values of LogP > -4.5 indicate that the compound can passively traverse the lipid bilayer and therefore is expected to be CNS-positive, for example.
  3. "Pathway" (GLMol) link provides the interactive 3D images of a compound moving across the membrane. An alternative visualization of the pathway can be obtained via PyMOL using “load XXXout.pdb, multiplex=1” command for the downloadable output coordinate file.
  4. The calculated permeability coefficients show a good correlation with experimental data for BLM (R2=0.87). Correlations between PerMM-calculated and experimental permeabilities for BBB and Caco-2 cell-based assays are more poor (R2 from 0.3 to 0.6 depending on compound class).


Please send any questions or requests to Andrei Lomize (