Some questions about docking (version xtb-6.6.0)

[codecollider]

Dear all,

I´m currently using the docking algorithm of xtb to get structures of cyclodextrin and POM adducts (1 POM and 1 to several cyclodextrins) as CREST takes too much time for these systems. To generate higher aggregates I first dock the first cyclodextrine to the POM and the structure looks reasonably good and is more stable than any "manually" generated structure. The problem comes in when I try to add the second cyclodextrin.

Shown on the left is what I expect and shown on the right is the outcome from xtb (I know that the left structure is more stable than the right). The POM is schematically drawn in blue and the cyclodextrin in orange.

--> Link to imgur with schematic drawing:

i.imgur.com/XiUVvRl.jpg

I assume that due to the limited number of starting orientations of the POM*CD unit and the CD unit the expected geometry is not found. Does this make any sense? Is there any parameter I can adjust to have more and possibly better resolved starting orientations? I tried also directed docking by defining prefered O atoms on the POM. It is still running, but from what I see it does not help.

Another thing that concerns me is that I get a lot of NaN energies in the genetic optimization algorithm (see below as an example output).

The command I used to execute the docking is:

xtb dock mol1_chrg_min6.xyz mol2_chrg_neutral.xyz --gfn 1 --maxgen 20

mol1_chrg_min6.xyz = POM*CD
mol2_chrg_neutral.xyz = CD

Charges I defined in .CHRG.

I also wonder were the temporary files are stored when doing a docking run. I have a opt_tmp file which has 0 bytes and one final_opt.xyz file that only includes the most recent calculation.

Thanks for any help!
Marco

`Precomputation of electronic porperties
For Molecule 1
Successful
For Molecule 2
Successful
LUMO energy 1 (read) : 5.506
HOMO energy 1 (read) : 0.500
LUMO energy 2 (read) : -3.735
HOMO energy 2 (read) : -11.952

Energy of cold fusion in kcal/mol: NaN

==============================================
| Starting Energy Screening |

Fast Mode selected (recommended)
If ATM term should be included, use -atm option.

Method for final opts. : gfn1

of genetic optimizations: 20

of parents : 100

of final geo. opts. : 15

Rare gas grid step size : 2.50
ang step size /deg : 45.00

angular grid points : 512

Performing stack search
Performing angular search

initialization done
Total gfn1 energy molecule 1: -678.2371515989
Total gfn1 energy molecule 2: -332.9670695220

---

Grid based energy screening

probe RG points :115200

Best rare gas probe energy/kcal : -4.53
+0.1 charged probe energy/kcal: -25.25
-0.1 charged probe energy/kcal: -5.37

Starting stack search
Grid points: 56000
lowest found /kcal : NaN

Starting angular search
Grid points:33792

Interaction energy of lowest structures so far in kcal/mol:
-177.10
-167.69
-150.72
-121.80
-87.02
-80.68
-69.20
-66.59
-64.35
-62.58

---

genetic optimization algorithm

cycle Eint/kcal/mol average Eint
1 -210.6 NaN
2 -101.0 NaN
3 NaN NaN
4 NaN NaN
5 NaN NaN
6 NaN NaN
7 NaN NaN
8 NaN NaN
9 NaN NaN
10 NaN NaN
11 NaN NaN
12 NaN NaN
13 NaN NaN
14 NaN NaN
15 NaN NaN
16 NaN NaN
17 NaN NaN
18 NaN NaN
19 NaN NaN
20 NaN NaN

Optimizing 15 best structures with gfn1
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15

---

  Interaction energies

---

Attention: monomers are not optimized
Interaction energies are not physical

Lowest Interaction Energy: -198.30 kcal/mol

E_int (kcal/mol)

1 -198.30
2 -197.78
3 -194.85
4 -194.80
5 -193.43
6 -191.42
7 -188.28
8 -187.52
9 -185.28
10 -183.50
11 -143.53
12 -65.09
13 -7.35
14 -4.40
15 83.01

---

• finished run on 2023/03/08 at 18:20:11.249

---

total:

• wall-time: 2 d, 1 h, 3 min, 46.002 sec
• cpu-time: 1 d, 18 h, 12 min, 49.400 sec
• ratio c/w: 0.860 speedup
  LMO Computation:
• wall-time: 0 d, 0 h, 0 min, 38.370 sec
• cpu-time: 0 d, 0 h, 0 min, 28.360 sec
• ratio c/w: 0.739 speedup
  SP:
• wall-time: 0 d, 0 h, 0 min, 0.263 sec
• cpu-time: 0 d, 0 h, 0 min, 1.099 sec
• ratio c/w: 4.178 speedup
  Searching:
• wall-time: 0 d, 0 h, 4 min, 24.476 sec
• cpu-time: 0 d, 0 h, 19 min, 43.051 sec
• ratio c/w: 4.473 speedup
  ANC optimizer:
• wall-time: 2 d, 0 h, 58 min, 36.656 sec
• cpu-time: 1 d, 17 h, 52 min, 7.012 sec
• ratio c/w: 0.855 speedup

normal termination of xtb
`

[cplett]

Hi Marco,
The NaNs should indeed not occur. I will try to reproduce and fix this, but it would be best if you could provide me with the structures or, if this is not possible, something similar.
Regarding the unexpected outcome of the second POM addition: I would at least expect that it is oriented parallel to the first one on the opposite side of the CD. This might actually be a bug indicated by the NaNs. Moreover, it could be a result of an inaccurate xTB-IFF energy or the structure screening that is done with the rigid monomers. For additional output, the --verbose flag can be used.
Maybe it would help to define the complete second POM for the directed docking to increase the attractive interaction also to the first POM.