Post-docking rescoring & pose analysis

An AMBER GPU rescoring pipeline for docked poses, with the scoring and analysis tools around it: minimise/rescore each pose with pmemd.cuda, then recompute ligand / binding-site RMSD and Fnat (fraction of native contacts), rank the poses, and summarise how the receptor-conformation ensemble performs. Works downstream of AutoDock, Vina and GNINA docking.

No absolute paths are hard-coded: tool locations default to $HOME/... in the pipeline scripts - edit them for your environment.

---

Requirements

• AmberTools + Amber (pmemd.cuda, cpptraj) for rescoring
• VMD for RMSD / Fnat scoring
• MGLTools, AutoDock/Vina/GNINA for the redocking (opt1) steps
• Python 3 (numpy, pandas) for the merge / population / analysis scripts

---

Pipeline

docked poses ─▶ full_pipeline_gpu*.sh ─▶ AMBER GPU min/rescore (pmemd.cuda)
                                              │
        reorder_ref_to_<engine>.tcl ─┐        ▼
        calc_bs_lig_rmsd_<engine>.sh ─┴▶ ligand / BS RMSD  ┐
        calc_fnat*.tcl              ─────▶ Fnat            ┼▶ merge_rmsds.py
                                                           ┘        │
                              cluster_ranking.py / population_*.py ◀─┘

---

Scripts by group

Rescoring pipelines

Script	Purpose
full_pipeline_gpu.sh	Full AMBER GPU (pmemd.cuda) rescoring of docked poses
full_pipeline_gpu_apo.sh	Same, for apo/unbound-receptor docking
full_pipeline_gpu_opt1_autodock.sh / _vina.sh / _gnina.sh	AMBER "opt1" rescoring + engine-specific --local_only redocking
run_vina_X_opt1.sh / run_gnina_X_opt1.sh	Config wrappers for the Vina / GNINA opt1 redocking
generate_opt1_pdb.sh	Build com_opt1.pdb complexes from the AMBER opt1.rst7 outputs

Scoring — Fnat & RMSD

Script	Purpose
calc_fnat.tcl	Fnat of one complex (ref.pdb vs complex_rec.pdb)
calc_fnat_all-poses.tcl	Fnat of every pose in a concatenated concat.pdb
calc_bs_lig_rmsd_autodock.sh / _vina.sh / _gnina.sh / _autodock-apo.sh	Binding-site + ligand RMSD, per docking engine
calc_RMSD_lig-complex.tcl	VMD ligand / complex RMSD template
cpptraj_rmsd.sh	cpptraj RMSD driver
reorder_ref_to_autodock.tcl / _vina.tcl / _gnina.tcl / _autodock-apo.tcl	Reorder reference atoms to each engine's atom order (for correct RMSD)

Fnat scripts consolidated. The original per-ligand Fnat scripts (ADP, AMP, AP5, B4P, G5P, GP5, …) are replaced by the two calc_fnat*.tcl above - set ligname (and optionally ligresid) at the top instead of using a separate file per ligand.

Ranking, population & analysis

Script	Purpose
merge_rmsds.py / merge_rmsds-apo.py	Merge the RMSD tracks into one table
filtered-rmsd-2.5.sh	Keep poses under an RMSD threshold (2.5 Å)
cluster_ranking.py	Rank cpptraj clusters by average top-N score
population_HM.sh	Fraction of receptor models giving a near-native pose (BS-RMSD < cutoff)
population_rmsd-energy.py	Population analysis over RMSD vs energy
analyze_reps.py	Analyse per-cluster representative poses

Docking / prep helpers

Script	Purpose
docking.sh	AutoDock docking driver (MGLTools + AutoGrid)
bs_dim_center.tcl	Binding-site box center + dimensions
create_concat_autodock.tcl	Concatenate AutoDock poses into a multi-MODEL PDB
prepare_ref_gnina.sh	Prepare the reference for GNINA scoring
fix_gnina_lig.py	Fix GNINA ligand output for downstream tools

---

License

MIT - see LICENSE.