Molecular Docking Basics

Lesson 3 of 12 · 11 min

Getting Structures from the Protein Data Bank

In docking, the receptor structure sets a ceiling on everything downstream: dock into a poorly chosen structure and you get confident-looking scores that mean nothing. So before running any calculation, you fetch candidate structures from the Protein Data Bank and read their quality metrics to pick the best one. This lesson works through that process using entry 1HSG, HIV-1 protease bound to the inhibitor indinavir and a long-standing docking benchmark.

Searching the RCSB PDB

The RCSB PDB at rcsb.org hosts every publicly released macromolecular structure, each keyed by a four-character ID such as 1HSG. Most entries are offered in two coordinate formats: the legacy .pdb format, which packs one atom per fixed-column line, and the newer mmCIF (.cif) format, a tag-value layout with no fixed-column size limits. Small and medium proteins are convenient as .pdb and nearly every structural tool reads it, but very large complexes exceed the fixed-column PDB format and are released only as .cif, so it pays to be comfortable with both.

Reading Quality Metrics

Four properties decide whether a structure is worth docking into: resolution in angstroms, where lower means sharper detail and side-chain and ligand positions blur above roughly 2.5 A, and the R-work and R-free factors, which measure how well the refined model reproduces the observed diffraction data. Good crystal structures typically sit in the 0.15 to 0.25 range for both, and a large gap between R-free and R-work warns of an overfit model. Missing residues, listed in REMARK 465 records, mark disordered regions the crystallographer could not build and matter most when they fall in your binding site, while a co-crystallized ligand sitting in that site confirms it is real and gives you a reference pose to validate your docking against.

Tip: For a docking receptor, prefer resolution at or below about 2.5 A, an R-free close to the R-work, no missing residues in the binding pocket, and ideally a ligand already bound in that pocket. 1HSG scores well at 2.00 A with an R-work of 0.166 and indinavir in the active site; being a 1995 structure it predates routine R-free reporting, so that field reads NULL.

Fetch and Inspect 1HSG

bash
# download HIV-1 protease + indinavir in PDB format (use wget if you prefer)
curl -O https://files.rcsb.org/download/1HSG.pdb
# the same structure in mmCIF: curl -O https://files.rcsb.org/download/1HSG.cif

grep '^HEADER' 1HSG.pdb
grep '^REMARK   2 RESOLUTION' 1HSG.pdb
grep 'R VALUE' 1HSG.pdb | grep -v NULL
HEADER    HYDROLASE (ACID PROTEINASE)             31-MAR-95   1HSG
REMARK   2 RESOLUTION.    2.00 ANGSTROMS.
REMARK   3   R VALUE            (WORKING SET) : 0.166
bash
# the bound ligand is stored as HETATM records; MK1 is indinavir
grep '^HETATM' 1HSG.pdb | grep ' MK1 ' | head -4
grep '^FORMUL' 1HSG.pdb

# crystallographic waters are HETATM records named HOH
grep '^HETATM' 1HSG.pdb | grep ' HOH ' | wc -l
HETATM 1517  N1  MK1 B 902       9.280  23.763   3.004  1.00 28.25           N
HETATM 1518  C1  MK1 B 902       9.498  23.983   4.459  1.00 30.30           C
HETATM 1519  C2  MK1 B 902      10.591  24.905   4.962  1.00 27.27           C
HETATM 1520  C3  MK1 B 902      10.591  24.864   6.466  1.00 28.85           C
FORMUL   3  MK1    C36 H47 N5 O4
FORMUL   4  HOH   *127(H2 O)
     127

Try it yourself: Fetch the estrogen receptor structure 3ERT the same way with curl -O https://files.rcsb.org/download/3ERT.pdb. Grep its HEADER for the classification and its REMARK 2 record for the resolution, then list its non-water HETATM records to find the bound ligand's three-letter code. Is 3ERT higher or lower resolution than 1HSG, and unlike 1HSG does it report an R-free value?