ADCP 1.1 documentation

Note

The documentation for this version is under construction. For now it gives a few example runs that show the new capabilities of version 1.1.

For general advice (how many runs and steps, which starting conformation, clashes and caps in the output), see the ADCP FAQ; it applies to version 1.1 too.

To run these examples, download the receptor and the ligand for PDB 3Q47.

Create a target file

This command computes the target file 3Q47.trg:

agfr -r 3Q47_recH.pdbqt -l 3Q47_pepH.pdbqt -asv 1.1 -o 3Q47

agfr runs AutoGridFR, which computes target files.

Option	Meaning
`-r`	the receptor, here as a PDBQT file
`-l`	the ligand, here the crystallographic peptide. When a ligand is given, agfr defines the docking box as the ligand’s bounding box with 4 Å padding.
`-asv 1.1`	compute the preferred starting ligand positions in the box with AutoSite 1.1, rather than 1.0, which is meant for small molecules
`-o 3Q47`	name of the target file to write (`3Q47.trg`)

Console output

processing: 3Q47_recH.pdbqt
loading ligand: 3Q47_pepH.pdbqt

set box using ligand
    Box center:    16.052     9.415     5.617
    Box length:    20.250    21.000    21.000
    Box size  :        54        56        56
    padding   :     4.000
    spacing   :     0.375

identifying pockets using AutoSite ....
Scanning at: -0.36 -0.792 -0.6
0 #: 0
Scanning at: -0.315 -0.6930000000000001 -0.525
clust.| Energy| # of |Rad. of | energy |   bns    |score   |
number|       |points|gyration|per vol.|buriedness|v*b^2/rg|
------+-------+------+--------+--------+----------+--------|
    1   -33.10    59    2.50    -0.56     0.729     12.54 
Scanning at: -0.27 -0.5940000000000001 -0.44999999999999996
clust.| Energy| # of |Rad. of | energy |   bns    |score   |
number|       |points|gyration|per vol.|buriedness|v*b^2/rg|
------+-------+------+--------+--------+----------+--------|
    1   -46.45    88    2.95    -0.53     0.772     17.78 
    2   -34.12    68    2.32    -0.50     0.756     16.79 
    3   -46.16    84    2.74    -0.55     0.715     15.65 
Scanning at: -0.22499999999999998 -0.495 -0.375
clust.| Energy| # of |Rad. of | energy |   bns    |score   |
number|       |points|gyration|per vol.|buriedness|v*b^2/rg|
------+-------+------+--------+--------+----------+--------|
    1   -86.02   169    3.58    -0.51     0.741     25.87 
    2   -66.14   140    3.26    -0.47     0.664     18.92 
    3   -73.08   162    4.53    -0.45     0.689     16.97 
    found 3 pocket(s)

    pocket|  energy | # of |Rad. of | energy |   bns    | score  
    number|         |points|gyration|per vol.|buriedness|v*b^2/rg
    ------+---------+------+--------+--------+----------+---------
        1   -102.22   260    4.15     -0.39      0.82      41.58
        2    -83.99   236    4.02     -0.36      0.77      34.84
        3    -92.58   265    4.92     -0.35      0.78      32.82
    merging clusters ...
done. got 107 fill Points, in 1.80 (sec)

setting map types using: all to ['H', 'HD', 'HS', 'C', 'A', 'G', 'GA', 'J', 'Q', 'N', 'NA', 'NS', 'OA', 'OS', 'SA', 'S', 'P', 'F', 'Cl', 'Br', 'I', 'Mg', 'Ca', 'Mn', 'Fe', 'Zn', 'Z']

computing maps for center=(16.052 9.415 5.617) size=(20.250 21.000 21.000) dims=(54 56 56) ...
    106 points inside the box

    maps computed in 1.93 (sec)
making target file 3Q47.trg ...done.
    done. 4.30 (sec)

Run ADCP with modified amino acids

adcp -T 3Q47.trg -s "n<BHD>pis&dvd" -N 2 -n 2000 -L swiss -w testDir -o testNSAAandD

In version 1.1 the peptide sequence can contain non-natural amino acids, written as [&]x<NAME>:

The optional & makes the amino acid D instead of the default L chirality.
x is the coarse bead that represents the side chain. Any one-letter code of a natural amino acid can be used here.
The modified side chain’s name goes between chevrons, <>.

In the example, the first amino acid is represented as an asparagine in the coarse model, and an asparagine derivative side chain (BHD) from the Swiss side-chain library is used for the full-atom representation. The peptide also contains a D-aspartic acid at position 5.

Tip

The modified side chain is loaded from a library file, here the Swiss library selected with -L swiss.

This example runs only 2 Monte Carlo searches (-N 2) of 2000 evaluations each, so that it finishes quickly; it is not meant to find real docked poses. It writes two files into the working directory testDir (-w):

testNSAAandD_out.pdb: a multi-model PDB file with the docked poses
testNSAAandD_summary.dlg: the docking log

Console output

Inflating target file 3Q47.trg
performing MC searches with: /home/xxxxxxx/micromamba/envs/adcpsuite/lib/python3.7/site-packages/ADCP/CrankiteAD_Linux-x86_64_1.1 
target data from file: 3Q47.trg
job name: testNSAAandD, summary file testNSAAandD_summary.dlg, docked poses: testNSAAandD_out.pdb
Detected 8 cores, using 8 cores.
using system rotamer libraries swiss from /home/xxxxxxx/micromamba/envs/adcpsuite/lib/python3.7/site-packages/ADCP/data/rotamers
Performing 2 MC searches using 2000 evals each using a random seed.
Performing search (2 ADCP runs with 2000 steps each) ...
0%   10   20   30   40   50   60   70   80   90   100%
|----|----|----|----|----|----|----|----|----|----|
Docking performed in 2.03 seconds, i.e. 0 hours 00 minutes 02.030427 seconds 
bestEnergies [12.8321, 11.8518] 
bestEnergy in run 2 11.851800 (0)
Analyzing results ....
concatenating trajectories for run with best energy < 31.851800
concatenated 2 trajectories
Clustering MC trajectories based in contacts using cutoff: 0.800000
finished calculating neighbors for 22 poses with 0.0 seconds
mode |  affinity  | ref. | clust. | rmsd | energy | best |
     | (kcal/mol) | fnc  |  size  | stdv |  stdv  | run  |
-----+------------+------+--------+------+--------+------+
   1          7.0      0.0       2      NA      NA    015 
   2          7.6      0.0       2      NA      NA    008 
   3         12.5      0.0       1      NA      NA    009 
Calculations completed 2.12 seconds, i.e. 0 hours 00 minutes 02.030427 seconds 
MC search command: cd testDir/testNSAAandD; /home/xxxxxxx/micromamba/envs/adcpsuite/lib/python3.7/site-packages/ADCP/CrankiteAD_Linux-x86_64_1.1 -t 2 "n<BHD>pis&dvd" -L swiss -T /home/xxxxxxx/ADCP_examples/testDir/testNSAAandD/3Q47 -r 1x2000 -p Bias=NULL,external=5,constrains,1.0,1.0,Opt=1,0.25,0.75,0.0 -s 930583 -o run_1.pdb  
seed: []
removing working folder testDir/testNSAAandD

OpenMM minimization

ADCP 1.1 can minimize docked poses with OpenMM and re-rank them. These examples assume you already know the standard ADCP options.

Re-ranking poses of standard amino acids

The option -nmin turns on OpenMM minimization of poses, followed by re-ranking. This example docks the peptide grctksicfpd into the receptor from 3Q47.trg and minimizes the top 5 docked solutions with OpenMM. The peptide is cyclic: -cyc closes it through its backbone (head to tail), while -cys would close it through a disulfide bond between two cysteines.

adcp -O -T 3Q47.trg -s "grctksicfpd" -cyc -N 5 -n 100000 -w testDir1 -o example1 -nmin 5

Console output

------------------------------------------------------------------
OpenMM minimization flag detected. This step takes more time than non-minimization calculations.

DECLARATION (V1.1.0 build 15):
a: Support for OpenMM Minimization is still under development.
b: Currently, it supports docking with "-rmsd 0" flag.
c: Current version provides docking supports for peptides containing ~400 (L and D) NSTs and openMM support for 182 (182x2 for D and L) NSTs. 
d: Other unidentified non-standard amino acids can either be replaced by similar amino acids (pdbfixer v1.8), or if pdbfixer does not identify a non-standard amino acid, it can be replaced by ALA.
e: Added support for external parameter input for non-standard amino acids.(from build 9)
   
Inflating target file 3Q47.trg
performing MC searches with: /home/xxxxxxx/micromamba/envs/adcpsuite/lib/python3.7/site-packages/ADCP/CrankiteAD_Linux-x86_64_1.1 
target data from file: 3Q47.trg
job name: example1, summary file example1_summary.dlg, docked poses: example1_out.pdb
Detected 8 cores, using 8 cores.
Performing 5 MC searches using 100000 evals each using a random seed.
Performing search (5 ADCP runs with 100000 steps each) ...
0%   10   20   30   40   50   60   70   80   90   100%
|----|----|----|----|----|----|----|----|----|----|
Docking performed in 22.04 seconds, i.e. 0 hours 00 minutes 22.042032 seconds 
bestEnergies [-4.91352, -0.869985, -0.258118, -3.55476, -3.64629] 
bestEnergy in run 1 -4.913520 (0)
Analyzing results ....
concatenating trajectories for run with best energy < 15.086480
concatenated 5 trajectories
Clustering MC trajectories based in contacts using cutoff: 0.800000
finished calculating neighbors for 125 poses with 0.6 seconds
mode |  affinity  | ref. | clust. | rmsd | energy | best |
     | (kcal/mol) | fnc  |  size  | stdv |  stdv  | run  |
-----+------------+------+--------+------+--------+------+
   1         -2.9      0.0      12      NA      NA    018 
   2         -2.3      0.0       6      NA      NA    020 
   3         -2.3      0.0       2      NA      NA    019 
   4         -2.2      0.0      16      NA      NA    113 
   5         -2.1      0.0       9      NA      NA    089 
   6         -1.7      0.0       8      NA      NA    080 
   7         -1.3      0.0       1      NA      NA    021 
   8         -1.1      0.0      11      NA      NA    072 
   9         -1.0      0.0       2      NA      NA    106 
  10         -0.9      0.0       3      NA      NA    003 
  11         -0.5      0.0       7      NA      NA    034 
  12         -0.5      0.0       4      NA      NA    061 
Calculations completed 22.77 seconds, i.e. 0 hours 00 minutes 22.042032 seconds 
MC search command: cd testDir1/example1; /home/xxxxxxx/micromamba/envs/adcpsuite/lib/python3.7/site-packages/ADCP/CrankiteAD_Linux-x86_64_1.1 -t 2 "grctksicfpd" -T /home/xxxxxxx/ADCP_examples/testDir1/example1/3Q47 -r 1x100000 -p Bias=NULL,external=5,constrains,1.0,1.0,external2=4,constrains,1.0,1.0,Opt=1,0.25,0.75,0.0 -s 390427 -o run_1.pdb  
seed: []
Minimizing docked poses ....
Rescoring clustered poses using OpenMM ..
OpenMM minimization settings: Environment="in-vacuo"; Max_itr=5
OpenMM minimization settings: Cyclizing by backbone
From total 44 models, minimizing top 5 ...

OMM Energy: Working on #1 of 5 models
Cyclizing peptide (last chain) by adding a bond between GLY1@N and ASP11@C.
OMM Energy: E_Complex =  -1201.00; E_Receptor =  -1586.36; E_Peptide  =    441.69
OMM Energy: dE_Interaction =    -56.33; dE_Complex-Receptor =    385.36

OMM Energy: Working on #2 of 5 models
Cyclizing peptide (last chain) by adding a bond between GLY1@N and ASP11@C.
OMM Energy: E_Complex =  18295.22; E_Receptor =   1122.39; E_Peptide  =  17239.90
OMM Energy: dE_Interaction =    -67.07; dE_Complex-Receptor =  17172.83

OMM Energy: Working on #3 of 5 models
Cyclizing peptide (last chain) by adding a bond between GLY1@N and ASP11@C.
OMM Energy: E_Complex =  -1721.60; E_Receptor =  -1855.51; E_Peptide  =    231.76
OMM Energy: dE_Interaction =    -97.86; dE_Complex-Receptor =    133.91

OMM Energy: Working on #4 of 5 models
Cyclizing peptide (last chain) by adding a bond between GLY1@N and ASP11@C.
OMM Energy: E_Complex =  -1387.83; E_Receptor =  -1685.81; E_Peptide  =    327.20
OMM Energy: dE_Interaction =    -29.22; dE_Complex-Receptor =    297.98

OMM Energy: Working on #5 of 5 models
Cyclizing peptide (last chain) by adding a bond between GLY1@N and ASP11@C.
OMM Energy: E_Complex =  -1784.56; E_Receptor =  -1940.05; E_Peptide  =    234.73
OMM Energy: dE_Interaction =    -79.25; dE_Complex-Receptor =    155.48

OMM Ranking:REARRANGING output poses using OpenMM (Ecomplex -Ereceptor) energy
OMM Ranking:                     +<-------OMMscore-------->+<-----------AutoDock CrankPep Scores-------------->+
OMM Ranking:-------+------+------+------------+------------+------------+------+--------+------+--------+------+
OMM Ranking: Model | Rank | Rank | E_Complex  |  E_Complex |  affinity  | ref. | clust. | rmsd | energy | best |
OMM Ranking: #     |OpenMM| ADCP |-E_Receptor |-E_rec-E_pep| (kcal/mol) | fnc  |  size  | stdv |  stdv  | run  |
OMM Ranking:-------+------+------+------------+------------+------------+------+--------+------+--------+------+
OMM Ranking:      1      1      3      133.9        -97.9          -2.3      0.0       2      NA      NA    019
OMM Ranking:      2      2      5      155.5        -79.2          -2.1      0.0       9      NA      NA    089
OMM Ranking:      3      3      4      298.0        -29.2          -2.2      0.0      16      NA      NA    113
OMM Ranking:      4      4      1      385.4        -56.3          -2.9      0.0      12      NA      NA    018
OMM Ranking:      5      5      2    17172.8        -67.1          -2.3      0.0       6      NA      NA    020
OMM Ranking:-------+------+------+------------+------------+------------+------+--------+------+--------+------+
removing working folder testDir1/example1

Besides the default ADCP output, this command writes:

example1_omm_rescored_out.pdb: the minimized and re-ranked top solutions (receptor + peptide)
testDir1/example1_omm_amber_parm/: Amber topology and coordinate files for each re-ranked solution, for further analysis or free-energy calculations with molecular dynamics or other molecular mechanics methods

Additional options

The number of minimization steps (-nitr), the solvent environment (-env) and the output order of re-ranked solutions (-dr) control the minimization. For example:

adcp -O -T 3Q47.trg -s "grctksicfpd" -cyc -N 5 -n 100000 -w testDir1 -o example1 -nmin 5 -nitr 500 -env implicit -dr -reint

Option	Meaning
`-nitr 500`	500 minimization steps for each docked pose
`-env implicit`	minimize in implicit solvent (default: vacuum)
`-dr`	report minimized solutions ordered by docking score, best to worst (default: OpenMM ranking score)
`-reint`	rank by interaction energy (E_complex − E_receptor − E_peptide) instead of (E_complex − E_receptor)

Minimizing poses that are already docked

Post-docking minimization (-pdmin) applies OpenMM minimization to poses that were already docked. ADCP skips docking and minimizes the poses from the earlier output file. Use the same target file (-T), job name (-o) and sequence (-s) as in the docking run.
Resume a failed minimization (-resumemin) continues an interrupted minimization and re-ranking. It works only together with -pdmin and without overwriting (-O).

Non-standard amino acids

The default force-field parameters for OpenMM minimization lack most non-standard amino acids. ADCP therefore uses the “sannerlab” OpenMM parameter set for 182 non-standard amino acids and the “swiss” set for 20 non-terminal ones. These sets are designed for the Swiss non-standard amino acids and need proper licensing from the developer of the SwissSideChain database. The full list of supported residues is in ADCP/data/openMMff/AVAILABLE_PARAMETER.dat.

When no OpenMM parameters exist for a non-standard amino acid, ADCP uses pdbfixer (v1.8) to replace it with a standard one. If pdbfixer cannot replace it, the residue is mutated to alanine for the OpenMM re-ranking.

Option	Meaning
`-F`	which OpenMM parameter sets to use from `ADCP/data/openMMff` (`swiss`, `sannerlab` or `none`); `none` turns off all system default sets
`-f`	user-supplied OpenMM parameters for new or unlisted non-standard amino acids (see below)

-f needs three files in the same directory with the same stem. For -f ./XXXX_ff.xml the program expects a force-field file ./XXXX_ff.xml, a bond definition file ./XXXX_residues.xml and a hydrogen definition file ./XXXX_hydrogen_def.xml.

Non-canonical amino acids

ADCP 1.1 docks peptides that contain non-natural amino acids (NSAAs). Each NSAA has a name (for example PTR), and ADCP reads its side-chain parameters from library files given on the command line (-l and -L). ADCP ships with a library for the natural amino acids (always loaded) and a library of fluorinated amino acids.

We also developed parameters for all amino acids of SwissSideChain. Its license will be replaced by CC-BY 4.0, and we have been granted the right to distribute the ADCP library for this set of 209 amino acids in D and L forms.

The library covers 182 of the 209 SwissSideChain amino acids, in both L and D chirality. This lets ADCP handle every non-natural amino acid currently found in the Protein Data Bank. To use peptides with NSAAs at any position in OpenMM, we:

Computed partial charges. SwissSideChain’s charges were designed for NSAAs away from the peptide termini, so we computed charges for terminal and non-terminal forms with the R.E.D. server (RESP ESP charge Derive), using GAMESS, following this tutorial. Each form was capped with ACE and/or NME.
Found force-field parameters. parmchk2 from AmberTools identified the Amber parameters needed (bonds, angles, dihedrals, impropers, electrostatics, van der Waals) from the R.E.D. mol2 files, written as an frcmod file.
Built FFXML files. ParmEd converted the frcmod and mol2 files into OpenMM FFXML files for each variant. These were merged into one file that OpenMM can load. ADCP/data/openMMff/AVAILABLE_PARAMETER.dat lists the residues it covers: 182 NSAAs in their internal, N-terminal and C-terminal forms.

To use these parameters in OpenMM we also changed how each structure is loaded, cleaned and simulated:

Kept pdbfixer from removing or swapping NSAAs, by loading the residue definition file and editing pdbfixer.nonstandardResidues.
Added OXT to C-terminal NSAAs. pdbfixer adds hydrogens with Modeller and handles N-terminal residues correctly, but not terminal NSAAs.
Overrode the default hydrogen definitions, to add NSAA-specific ones or apply specific parameter sets to residues with identical names (for example ORN).
Gave residues with identical molecular structures (for example THR and ALO) their own residue templates.
Overrode PDB writing to handle 4-letter residue codes, with a modified PDB writer.