A vacated second ligand site acts as a decoy in redocking: systematic in P-glycoprotein, intermittent across eight other proteins

Redocking tests whether a docking protocol can recover a deposited pose: the ligand is removed and the program is asked to put it back. Standard preparation deletes every copy. When a structure contains two copies in one cavity, the receptor passed to the program was never observed empty. We measured the effect of that deletion in deposited human P-glycoprotein structures, with the search box centred on the deposited pose so that a failure cannot be blamed on site finding. Across 15 structures the median top-pose RMSD is 8.30 Å and 1 of 15 falls within 2.0 Å. Accuracy does not track the density support of the deposited ligand. That was the registered hypothesis, and it is not supported (Spearman ρ = −0.004, p = 0.9899). In the nine structures that carry a second copy, retaining the copy as rigid receptor moves the median from 8.26 to 3.46 Å and improves 8 of 9 (p = 0.0078). A carbon-only ghost of the same shape, which the scoring function sees as apolar volume but not as a hydrogen-bonding partner, agrees with the retained copy (p = 0.1289), while a box reduced to the same volume does not (p = 0.2969). The displaced poses lie in the vacated subsite (mean overlap 0.216 to 0.539, p = 0.0078). The comparison reproduces in rDock, which shares no code with smina: 6 of 7 structures improve, and the median falls from 8.35 to 3.67 Å (p = 0.0469). On eight unrelated proteins the same protocol does not support one direction (5 of 8, p = 0.3750), although trypsin moves from 8.59 to 0.50 Å. A scan of 37,596 PDB entries found 4,774 multi-copy sites across 929 proteins. A redocking number for a multi-copy structure should be reported both with the other copies removed and with them retained. Scientific Contribution. Published evidence for this preparation effect was one ligand in one structure. The comparison here covers nine two-copy P-glycoprotein structures, and a carbon ghost together with a volume-matched box separates apolar occlusion from polar ligand chemistry and from the volume searched. Repeated on eight other proteins, the same comparison does not support a consistent direction.

Authors

Institutions

Publication Details

Journal
ChemRxiv
Published
2026-11-16
DOI
https://doi.org/10.26434/chemrxiv.15012268/v1
Primary Topic
Computational Drug Discovery Methods
Type
preprint
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
preprint

A vacated second ligand site acts as a decoy in redocking: systematic in P-glycoprotein, intermittent across eight other proteins

Hayat Ullah, Moon-Il Joo, Hee-Cheol Kim, Md Ariful Islam Mozumder et al.
1 links
ChemRxiv
Computational Drug Discovery Methods
preprint

A vacated second ligand site acts as a decoy in redocking: systematic in P-glycoprotein, intermittent across eight other proteins

Hayat Ullah, Moon-Il Joo, Hee-Cheol Kim, Md Ariful Islam Mozumder, Muhammad Yaseen, Louisa Ayamga, Nada Mohamed Yassein Farih, Zeba khan, Khwaja Jawad, Ziaullah Khan
preprint en

Abstract

Redocking tests whether a docking protocol can recover a deposited pose: the ligand is removed and the program is asked to put it back. Standard preparation deletes every copy. When a structure contains two copies in one cavity, the receptor passed to the program was never observed empty. We measured the effect of that deletion in deposited human P-glycoprotein structures, with the search box centred on the deposited pose so that a failure cannot be blamed on site finding. Across 15 structures the median top-pose RMSD is 8.30 Å and 1 of 15 falls within 2.0 Å. Accuracy does not track the density support of the deposited ligand. That was the registered hypothesis, and it is not supported (Spearman ρ = −0.004, p = 0.9899). In the nine structures that carry a second copy, retaining the copy as rigid receptor moves the median from 8.26 to 3.46 Å and improves 8 of 9 (p = 0.0078). A carbon-only ghost of the same shape, which the scoring function sees as apolar volume but not as a hydrogen-bonding partner, agrees with the retained copy (p = 0.1289), while a box reduced to the same volume does not (p = 0.2969). The displaced poses lie in the vacated subsite (mean overlap 0.216 to 0.539, p = 0.0078). The comparison reproduces in rDock, which shares no code with smina: 6 of 7 structures improve, and the median falls from 8.35 to 3.67 Å (p = 0.0469). On eight unrelated proteins the same protocol does not support one direction (5 of 8, p = 0.3750), although trypsin moves from 8.59 to 0.50 Å. A scan of 37,596 PDB entries found 4,774 multi-copy sites across 929 proteins. A redocking number for a multi-copy structure should be reported both with the other copies removed and with them retained. Scientific Contribution. Published evidence for this preparation effect was one ligand in one structure. The comparison here covers nine two-copy P-glycoprotein structures, and a carbon ghost together with a volume-matched box separates apolar occlusion from polar ligand chemistry and from the volume searched. Repeated on eight other proteins, the same comparison does not support a consistent direction.

ChemRxiv
Inje University (KR)
Computational Drug Discovery Methods
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon