Writing the PDBQT files

What the two save buttons write, and how every choice is recorded.

The example below follows ADFR's own test of this entry. Starting from 6GGD, Biomol 2, the choices were:

check ZN'401 (under manual charges) into the receptor;
check EYB'402 (under ligands) in the lig. column;
mutate SER'96 to ILE.

Then receptor PDBQT and ligand PDBQT were pressed.

the two files written, with the log open — Each save is listed under **Files written** with a download link; the browser also downloads it at once. The **Log** shows what ADFR did: the side chains it completed, the mutation, the charged zinc.

The receptor

receptor PDBQT takes the receptor atoms of the current tab and runs ADFR's receptor preparation:

selenomethionines become methionines;
mutations are made and missing side chains built (best-scoring rotamer);
the alternate locations not chosen are dropped;
hydrogens are added with reduce;
Gasteiger charges are assigned with Open Babel, and manual charges are set on the ions;
AutoDock atom types are assigned, and non-polar hydrogens merged.

The file begins with REMARK 850 records of what was done; every record is explained in Provenance in the PDBQT files. For the session above:

REMARK 850 AGFR version 1.2dev
REMARK 850 AGFR source 6ggd.pdb
REMARK 850 AGFR biomolecule Biomol 2: chains B
REMARK 850 AGFR mutation B:SER:96:_ -> ILE
REMARK 850 AGFR side chain completed A:LYS:120:_
   …
REMARK 850 AGFR side chain completed B:LYS:291:_
REMARK 850 AGFR altlocs B:LYS:101:_ @B (0.52)
REMARK 850 AGFR altlocs B:SER:149:_ @C (0.42)
   …
REMARK 850 AGFR altlocs B:CYS:277:_ @A (0.57)
REMARK 850 AGFR protonated using reduce
REMARK 850 AGFR partial charge assigned using OpenBabel Gasteiger model
REMARK 850 AGFR end

The biomolecule line records the tab the receptor was written from and the chains of the asymmetric unit it uses. For an assembly built by crystal symmetry it also gives the number of copies, for instance REMARK 850 AGFR biomolecule Biomol 1: chains A B, 3 copies for 1H6W. ADapt adds this line, to the ligand files as well; the rest of each header is ADFR's.

Reading the REMARKs. The side chain completed lines list every completion queued in the file, including chains not in the current tab (chain A here, while Biomol 2 is chain B). The altlocs lines list only residues in the receptor, with the location used. This is how ADFR writes the header.

The protonated receptor (PDB)

Next to the PDBQT, the same save writes <id>_rec_H.pdb: the receptor as reduce protonated it, before ADFR merges the non-polar hydrogens and converts to PDBQT. It holds every hydrogen, polar and non-polar, at the positions reduce chose, and the same heavy atoms as the PDBQT. It is useful where a full-atom receptor is needed, and as the record of reduce's choices for histidines, asparagines and glutamines. For 1A3R (tab ASU and Biomol 1):

REMARK 850 AGFR version 1.2dev
REMARK 850 AGFR source 1a3r.pdb
REMARK 850 AGFR biomolecule ASU and Biomol 1: chains H L P
REMARK 850 AGFR protonated using reduce
REMARK 850 AGFR partial charge assigned using OpenBabel Gasteiger model
REMARK 850 AGFR end
USER  MOD reduce.3.23.130521 H: found=0, std=0, add=3340, rem=0, adj=160
   …
USER  MOD Set 1.1: L 198 HIS     :     no HD1:sc=   0.771  K(o=1.6,f=-4.4)
USER  MOD Set 3.1: L 190 ASN     :FLIP  amide:sc=   0.105  F(o=-0.34,f=0.18)
   …
ATOM      1  N   ASP L   1      79.656  63.382 102.827  1.00 46.17         A N
   …
ATOM      9  H1  ASP L   1      78.913  63.312 103.312  1.00 46.17         A H   new

The file starts with the PDBQT's REMARK 850 header, up to REMARK 850 AGFR end; everything after it is reduce's output, with the atoms numbered in order (below). Its USER MOD lines give, for each histidine, asparagine and glutamine whose hydrogens or orientation reduce decided, the decision (no HD1: protonated on NE2 only; FLIP: the amide or ring was flipped), a score, and a category: K keep, F flip, C clash, X uncertain. The same decisions are recorded in session.json (below).

Written as reduce writes it, numbered by ADapt. reduce gives every hydrogen it adds the atom serial number 0; ADapt numbers all atoms 1, 2, … in file order, so each has its own number. That is the only change to reduce's text. From the 100,000th atom on the number is hexadecimal (186a0, …), as ProDy writes it. The added hydrogens keep the word new after column 80; flipped atoms carry flip. Chain identifiers are those reduce saw: in an assembly built by symmetry, the copies are renamed 0, 1, …, as in the PDBQT (Known issues).

The ligand

ligand PDBQT takes the ligand atoms of the current tab. Each connected fragment becomes one ligand, prepared by ADFR: hydrogens added and Gasteiger charges assigned with Open Babel (pH 7.4), non-polar hydrogens merged, and a torsion tree built with an automatic root.

REMARK 850 file prepared by AGFR version 1.2dev
REMARK 850 source .pdb
REMARK 850 AGFR biomolecule Biomol 2: chains B
REMARK 850 AGFR ligand from B:EYB:402:_
REMARK 4 active torsions:
REMARK  status: ('A' for Active; 'I' for Inactive)
REMARK    0  A    between atoms: C10_7 and C18_14
   …
ROOT
HETATM    1  N1  EYB A 402     124.498 103.049 -42.674  1.00  0.00    -0.34  N

The chain letter. This EYB is chain B's; its coordinates are chain B's, but its atom records label it chain A, because the chain identifier is not kept through ADFR's Open Babel conversion (AGFRgui's file is the same). ADapt's REMARK 850 AGFR ligand from B:EYB:402:_ line records where it came from.

The files are named as AGFRgui names them: <id>_rec.pdbqt and <id>_lig.pdbqt, or <id>_lig_001.pdbqt, … when there are several fragments. ADapt adds <id>_rec_H.pdb to each receptor save. The browser downloads the receptor PDBQT, or the ligand file (a zip when there are several); Files written has a link for every file.

The session record

Every loaded file has a session folder (see Getting started). Its session.json holds everything needed to trace a file back to what produced it:

the input file, its size and md5, and where it came from (upload, or RCSB URL);
the tool (ADapt 0.1.0, its git commit), the Python version, the ADFR package (adfr-0.0.22) and the md5 of the ADFR source files that did the work;
every edit in order: tab switches, checkboxes, alternate locations, charges, side-chain choices, mutations;
every save: its folder, the files and their md5, and the mutations, alternate locations and charges it was made with;
for a receptor save, reduce's version and decisions, and the chain renames made for it.

"events": [
  {"what": "tab",    "name": "Biomol 2", "at": "2026-09-29T14:30:44"},
  {"what": "check",  "key": "B:ZN:401:_",  "column": "rec", "state": "checked", ...},
  {"what": "check",  "key": "B:EYB:402:_", "column": "lig", "state": "checked", ...},
  {"what": "mutate", "key": "B:SER:96:_",  "choice": "ILE", ...}
]

The reduce record of a receptor save, for the 1A3R receptor above:

"reduce": {
  "version": "reduce.3.23.130521",
  "chain_renames": {},
  "decisions": [
    {"residue": "L:HIS:198:_", "set": "Set 1.1", "decision": "no HD1", "score": 0.771,
     "category": "K", "detail": "o=1.6,f=-4.4"},
    {"residue": "L:ASN:190:_", "set": "Set 3.1", "decision": "FLIP  amide", "score": 0.105,
     "category": "F", "detail": "o=-0.34,f=0.18"},
    ...
  ]
}

decisions has one entry per histidine, asparagine and glutamine USER MOD line of _rec_H.pdb, with the residue written chain:name:number:insertion code as in the REMARKs. A residue reduce did not adjust (a NoAdj-H line, as for a histidine bound to a metal) is kept as {"raw": "<the line>"}. chain_renames maps each chain identifier given to a symmetry copy back to its segment and original chain, for instance "2": {"segment": "C", "chain": "A"} in 1H6W's Biomol 1; it is empty when nothing was renamed.

Checking against ADFR's command line

For the choices ADFR's command line can express, the receptor ADapt writes is byte-identical to the command line's, apart from the biomolecule line. The command line has no ligand selection, so the check uses a variant of the session above that keeps both ZN and EYB in the receptor (step 2 with the first box instead of lig.). That receptor was checked to be byte-identical to:

python -m ADFR.bin.runAGFR -r 6ggd.pdb --toPdbqt --bioMol 2 \
    --heteroToKeep "chid B resnum 402 401" --mutate "B:SER:96::ILE"

ADapt's test suite checks the same equivalence on ADFR's own test files. Run the command with the adcpsuite Python, with ADS_ROOT set to that environment. Use python -m rather than the agfr launcher, which splits option values that contain spaces.

← Mutating residues Provenance in the PDBQT files →