Table of Contents

• Introduction
• Repository Structure
• Benchmark entries
  • Dimeric complexes
  • Trimeric complexes
• Analysis
  • HADDOCK terms
  • RMSD terms
• Reference

Membrane protein complexes docking benchmark. This repository contains the entries of the Membrane Protein docking benchmark (V1.0). It also contains ProFit analysis scripts for the calculation of Interface- and Ligand-RMSD (I-/L-RMSD) values for all the entries of the benchmark. We have generated docking decoys for all the entries of the benchmark with the HADDOCK webserver. In addition to I- and L-RMSDs we have calculated van der Waals, Electrostatic and Desolvation energies and the Buried Surface Area (BSA) for all decoys of all complexes.

The repository structure is the following:

|   LICENSE
|   README.md
|   analysis
|   \   random_restraints
|   |   \   1k4c.txt
|   |   |   1m56.txt
|   |   /   ...
|   |   true_interface
|   |   \   1k4c.txt
|   |   |   1m56.txt
|   /   /   ...
|   structures
|   \   1k4c
|   |   \   1k4c_bound.pdb
|   |   |   1k4c_unbound.pdb
|   |   |   1k4c_unbound_A.pdb
|   |   |   1k4c_unbound_C.pdb
|   |   |   AC.irmsd
|   |   |   AC.izone
|   |   /   AC.lzone
|   |   1m56
|   |   \   1m56_bound.pdb
|   |   |   1m56_unbound.pdb
|   |   |   1m56_unbound_AB.pdb
|   |   |   1m56_unbound_CD.pdb
|   |   |   AB.irmsd
|   |   |   AB.izone
|   |   /   AB.lzone

The structures folder contains all the structure files in PDB format. Both reference (bound) and unbound structures are included. All structures have been modified so that chain identifiers and residue numbering is consistent.

Every subfolder in the structures folder (eg 1k4c) contains the following files:

• *_bound.pdb -> The reference structure for a given complex.
• *_unbound_*.pdb-> The unbound structures for a given complex.
• *_unbound.pdb -> The unbound structures optimally and independently superimposed on the reference chains.
• *.irmsd -> The backbond I-RMSD (in Å) of the unbound complex relative to the reference structure.
• *.izone -> ProFit-compatible I-RMSD calculation script.
• *.lzone -> ProFit-compatible L-RMSD calculation script.

The two subfolders in the analysis folder contain the results of the docking runs for two docking scenarios. One using only random restraints to drive the docking (random_restraints) and one using restraints derived from the interface of the bound complex (true_interface). Both folders contain 37 space-separated text files that are the results of the analysis of the docking run of a given complex.

The two tables below list the dimeric and trimeric entries respectively. The second column is the PDB id of the bound complex, the second and third (second through fourth for the trimeric table) list the origin of the unbound structures for that complex. The complexes that appear more than once in the benchmark are idenitified by the presence of an identifier in the complex_id field (eg 2r6g-TM).

The Category column refers to the nature of the complex. MS stands for complexes whose interface lies along the membrane-soluble region, TM stands for complexes whose interface lies within the membrane, Both stands for complexes whose interface lies in both membrane and soluble regions, Buried stands for complexes where one of the partners is buried in a transmembrane β-barrel and AB stands for complexes where one of the partners is an antibody-like domain.

The Composition column refers to the origin of the unbound structures of a complex. BB stands for Bound-Bound and means all partners have been extracted from the bound complex, UB stands for Bound-Unbound and means at least one of the partners has been extracted from an alternative structure and UU stands for Unbound-Unbound and means none of the partners originate in the bound complex.

The Difficulty column reflects the difficulty of a given complex with Bound cases having an I-RMSD of 0, Easy cases having an I-RMSD between 0 and 1Å, Intermediate cases having an I-RMSD between 1 and 2Å and Hard cases having an I-RMSD greater than 2Å.

All the RMSD values have been calculated with ProFit and all energy values along with Surface area by HADDOCK.

The following is an example of the results of the analysis for dimeric complex.

$ head 1k4c.txt
complex i-RMSD l-RMSD HS Eair Evdw Eelec Edesolv BSA stage pdb rank difficulty
1 0.959 2.102 -35.7645987 417.16 9.01013 -14.8071 -12.3443 1287.49 it0 1k4c 705 Easy
2 13.601 30.105 -10.0395301 1298.94 -1.60801 -4.26846 -6.80819 1193.62 it0 1k4c 6360 Easy
3 10.638 26.939 -25.8562489 1185.26 4.58911 -5.10454 -19.5606 1308.96 it0 1k4c 3083 Easy
4 6.757 20.617 -26.756805 990.199 12.5825 -9.88172 -11.7789 1512.4 it0 1k4c 2916 Easy
5 0.934 4.333 -31.571787 602.385 24.9723 -16.4337 -8.06686 1334.48 it0 1k4c 2137 Easy
6 14.473 29.379 -2.988335 1385.01 26.7835 -1.33664 -4.52393 1124.57 it0 1k4c 8595 Easy
7 14.348 29.247 -4.985707 1253.87 20.8563 -0.97943 -5.58424 1116.93 it0 1k4c 8029 Easy
8 10.826 26.383 -25.9056835 1206.93 -2.13535 -4.64953 -19.672 1363.21 it0 1k4c 3074 Easy
9 6.213 27.452 -10.763576 1411.04 24.4844 -2.1838 -14.0344 890.062 it0 1k4c 6111 Easy

and for a trimeric one

complex i-RMSD l-RMSD l-RMSD-sd HS Eair Evdw Eelec Edesolv BSA stage pdb rank difficulty
1 46.268 75.6035 1.8335 1790.72757 102074 718.818 -13.7373 7.13569 1901.6 it0 4fz0 8305 Intermediate
2 33.500 51.8455 1.1275 401.32159 38540.4 248.141 -9.81252 -51.7452 3421.41 it0 4fz0 393 Intermediate
3 36.232 58.523 17.595 673.51548 50304 375.559 53.4633 5.52919 3188.66 it0 4fz0 3527 Intermediate
4 47.686 62.758 6.711 636.45832 53861.1 356.872 -10.6338 31.3379 3474.35 it0 4fz0 3109 Intermediate
5 53.633 89.111 14.586 1552.34581 117606 412.29 -7.01079 6.8906 1238.89 it0 4fz0 7804 Intermediate
6 18.942 41.166 0.153 663.585429 53746.8 388.635 12.9086 -0.648221 2777.93 it0 4fz0 3401 Intermediate
7 38.178 61.3475 1.0615 579.84936 46221.8 172.666 42.2576 15.0654 3910.15 it0 4fz0 2494 Intermediate
8 34.095 62.5555 16.3635 833.98877 66967.1 250.811 5.05806 34.3765 3304.69 it0 4fz0 5141 Intermediate
9 49.997 91.598 8.271 2210.46723 89319.8 419.985 -7.88672 -13.6433 2024.06 it0 4fz0 8940 Intermediate

The values that have been calculated with HADDOCK are the Eair, Evdw, Eelec, Edesolv and BSA terms. HS stands for HADDOCK-SCORE which is a weighted sum of the pervious terms. Eair refers to the restraint energy, Evdw and Eelec are the non-bonded terms and are calculated with the OPLS force field, Edesolv stands for desolvation energy and BSA stands for Buried Surface Area.

The terms that are related to HADDOCK but are not included in the scoring of complexes are stage, and rank. Stage refers to one of three stages of a HADDOCK run:

• it0 - Rigid-body energy minisation
• it1 - Simulated annealing in torsional space
• itw - Flexible refinement in explicit solvent

Rank refers to the ranking of that particular structure by the HADDOCK scoring function.

The two RMSD terms are i-RMSD and l-RMSD. They stand for interface and ligand RMSD respectively. For the I-RMSD calculation we have calculated the RMSD of the backbone atoms of the interface. The interface is defined as the set of residues whose atoms are within 10Å of any atom of a partner. For the L-RMSD calculation we have superimposed the receptors (defined as the biggest partner) using all backbone atoms and calculated the displacement of the backbone atoms of the ligand (defined as the smallest partner). For the trimers, we have calculated the displacement of both partners (with regards to the first) and report their mean and standard deviation.

Those interested in reproducing the RMSD calculations can do so with the following ProFit commands

Using complex 1m56 as an example.

For the I-RMSD calculations:

profit -f AB.izone 1m56_bound.pdb 1m56_unbound.pdb

If ProFit is correctly installed on your system you should see the following output:

              PPPPP                FFFFFF ii   tt
              PP  PP               FF          tt
              PP  PP rrrrr   oooo  FF     ii  ttttt
              PPPPP  rr  rr oo  oo FFFF   ii   tt
              PP     rr     oo  oo FF     ii   tt
              PP     rr     oo  oo FF     ii   tt
              PP     rr      oooo  FF      ii   ttt

                  Protein Least Squares Fitting

                           Version 3.1

  Copyright (c) Dr. Andrew C.R. Martin, SciTech Software 1992-2009
          Copyright (c) Dr. Craig T. Porter, UCL 2008-2009

 Reading reference structure...
 Reading mobile structure...

 Starting script: 'AB.izone'
 Fitting structures...
 RMS: 0.572
 Finished script: 'AB.izone'

The line starting with RMS contains the I-RMSD value.

For the L-RMSD calculations:

profit -f AB.lzone 1m56_bound.pdb 1m56_unbound.pdb

Which should print:

              PPPPP                FFFFFF ii   tt
              PP  PP               FF          tt
              PP  PP rrrrr   oooo  FF     ii  ttttt
              PPPPP  rr  rr oo  oo FFFF   ii   tt
              PP     rr     oo  oo FF     ii   tt
              PP     rr     oo  oo FF     ii   tt
              PP     rr      oooo  FF      ii   ttt

                  Protein Least Squares Fitting

                           Version 3.1

  Copyright (c) Dr. Andrew C.R. Martin, SciTech Software 1992-2009
          Copyright (c) Dr. Craig T. Porter, UCL 2008-2009

 Reading reference structure...
 Reading mobile structure...

 Starting script: 'AB.lzone'
 Fitting structures...
 RMS: 0.560
 RMS: 0.000
 Finished script: 'AB.lzone'

In this case two RMSD values are calculated and printed to the terminal. The first is the RMSD of the backbone atoms of the receptor (in this case chain A). The second is the L-RMSD value. In this case it is 0 as chain B has been extracted from the bound complex for case 1m56.

When using this membrane complexes benchmark in any publication, please cite:

• P.I. Koukos, I. Faro, C.W. van Noort and A.M.J.J. Bonvin. A membrane protein complex docking benchmark. J. Mol. Biol. Advanced Online Publication (2018).