Unexpected high positive affinity score and a constant "ENERGY FROM SEARCH"

[AquifersBSIM]

I have docked 90,000 ligands and I have this unexpectedly very high positive affinity score. Other than that, if you look at the "ENERGY FROM SEARCH", it is a constant value throughout all the unexpected very high affinity score.

Here is the example

ENERGY FROM SEARCH: 340282346638528859811704183484516925440.0000
FINAL ENERGY:
Estimated Free Energy of Binding   : 24700.086 (kcal/mol) [=(1)+(2)+(3)+(4)]
(1) Final Intermolecular Energy    : 31197.938 (kcal/mol)
    Ligand - Receptor              : 31197.938 (kcal/mol)
    Ligand - Flex side chains      : 0.000 (kcal/mol)
(2) Final Total Internal Energy    : 1.936 (kcal/mol)
    Ligand                         : 1.936 (kcal/mol)
    Flex   - Receptor              : 0.000 (kcal/mol)
    Flex   - Flex side chains      : 0.000 (kcal/mol)
(3) Torsional Free Energy          : -6497.852 (kcal/mol)
(4) Unbound System's Energy        : 1.936 (kcal/mol)

mode |   affinity | dist from best mode
     | (kcal/mol) | rmsd l.b.| rmsd u.b.
-----+------------+----------+----------
   1    24700.0859     0.0000     0.0000

ENERGY FROM SEARCH: 340282346638528859811704183484516925440.0000
FINAL ENERGY:
Estimated Free Energy of Binding   : 8311.212 (kcal/mol) [=(1)+(2)+(3)+(4)]
(1) Final Intermolecular Energy    : 9525.896 (kcal/mol)
    Ligand - Receptor              : 9525.896 (kcal/mol)
    Ligand - Flex side chains      : 0.000 (kcal/mol)
(2) Final Total Internal Energy    : 1.443 (kcal/mol)
    Ligand                         : 1.443 (kcal/mol)
    Flex   - Receptor              : 0.000 (kcal/mol)
    Flex   - Flex side chains      : 0.000 (kcal/mol)
(3) Torsional Free Energy          : -1214.684 (kcal/mol)
(4) Unbound System's Energy        : 1.443 (kcal/mol)

mode |   affinity | dist from best mode
     | (kcal/mol) | rmsd l.b.| rmsd u.b.
-----+------------+----------+----------
   1    8311.2119     0.0000     0.0000

ENERGY FROM SEARCH: 340282346638528859811704183484516925440.0000
FINAL ENERGY:
Estimated Free Energy of Binding   : 1381.443 (kcal/mol) [=(1)+(2)+(3)+(4)]
(1) Final Intermolecular Energy    : 1704.479 (kcal/mol)
    Ligand - Receptor              : 1704.479 (kcal/mol)
    Ligand - Flex side chains      : 0.000 (kcal/mol)
(2) Final Total Internal Energy    : 2.270 (kcal/mol)
    Ligand                         : 2.270 (kcal/mol)
    Flex   - Receptor              : 0.000 (kcal/mol)
    Flex   - Flex side chains      : 0.000 (kcal/mol)
(3) Torsional Free Energy          : -323.037 (kcal/mol)
(4) Unbound System's Energy        : 2.270 (kcal/mol)

mode |   affinity | dist from best mode
     | (kcal/mol) | rmsd l.b.| rmsd u.b.
-----+------------+----------+----------
   1    1381.4429     0.0000     0.0000

I might be wrong here, but the structure and .pdbqt file for the highest scoring ligand looks fine.
Here is the .pdbqt for the highest scoring ligands:

REMARK  Name = PV-008503443598
REMARK  6 active torsions:
REMARK  status: ('A' for Active; 'I' for Inactive)
REMARK    1  A    between atoms: C_1  and  C_2
REMARK    2  A    between atoms: C_5  and  C_7
REMARK    3  A    between atoms: C_10  and  C_11
REMARK    4  A    between atoms: C_11  and  O_12
REMARK    5  A    between atoms: C_19  and  C_21
REMARK    6  A    between atoms: C_23  and  C_24
REMARK                            x       y       z     vdW  Elec       q    Type
REMARK                         _______ _______ _______ _____ _____    ______ ____
ROOT
ATOM      1  C   UNL     1      -0.822  -2.856   0.758  0.00  0.00    +0.000 C
ATOM      2  C   UNL     1      -0.552  -1.354   0.825  0.00  0.00    +0.000 C
ATOM      3  C   UNL     1      -1.464  -0.548  -0.121  0.00  0.00    +0.000 C
ATOM      4  N   UNL     1      -1.249   0.912   0.031  0.00  0.00    +0.000 N
ATOM      5  C   UNL     1      -2.162   1.849  -0.463  0.00  0.00    +0.000 C
ATOM      6  O   UNL     1      -1.860   2.999  -0.769  0.00  0.00    +0.000 OA
ATOM      7  C   UNL     1       0.157   1.350   0.070  0.00  0.00    +0.000 C
ATOM      8  C   UNL     1       1.042   0.468   0.970  0.00  0.00    +0.000 C
ATOM      9  N   UNL     1       0.852  -0.946   0.633  0.00  0.00    +0.000 N
ATOM     10  C   UNL     1       1.586  -1.488  -0.425  0.00  0.00    +0.000 C
ATOM     11  O   UNL     1       1.138  -2.289  -1.235  0.00  0.00    +0.000 OA
ENDROOT
BRANCH   5  12
ATOM     12  C   UNL     1      -3.575   1.449  -0.618  0.00  0.00    +0.000 A
ATOM     13  C   UNL     1      -4.160   1.497  -1.886  0.00  0.00    +0.000 A
ATOM     14  C   UNL     1      -5.501   1.136  -2.043  0.00  0.00    +0.000 A
ATOM     15  C   UNL     1      -6.265   0.732  -0.938  0.00  0.00    +0.000 A
ATOM     16  C   UNL     1      -5.662   0.733   0.325  0.00  0.00    +0.000 A
ATOM     17  F   UNL     1      -6.381   0.390   1.403  0.00  0.00    +0.000 F
ATOM     18  C   UNL     1      -4.328   1.087   0.499  0.00  0.00    +0.000 A
BRANCH  15  19
ATOM     19  C   UNL     1      -7.720   0.372  -1.057  0.00  0.00    +0.000 C
BRANCH  19  20
ATOM     20  O   UNL     1      -8.038  -0.044  -2.380  0.00  0.00    +0.000 OA
ATOM     21  H   UNL     1      -7.856   0.680  -2.999  0.00  0.00    +0.000 HD
ENDBRANCH  19  20
ENDBRANCH  15  19
ENDBRANCH   5  12
BRANCH  10  22
ATOM     22  C   UNL     1       3.009  -1.065  -0.566  0.00  0.00    +0.000 C
ATOM     23  C   UNL     1       4.060  -2.152  -0.586  0.00  0.00    +0.000 C
ATOM     24  C   UNL     1       3.998  -1.198   0.580  0.00  0.00    +0.000 C
BRANCH  24  25
ATOM     25  C   UNL     1       5.158  -0.245   0.769  0.00  0.00    +0.000 A
ATOM     26  C   UNL     1       6.381  -0.720   1.274  0.00  0.00    +0.000 A
ATOM     27  C   UNL     1       7.473   0.128   1.473  0.00  0.00    +0.000 A
ATOM     28  C   UNL     1       7.364   1.481   1.173  0.00  0.00    +0.000 A
ATOM     29  C   UNL     1       6.166   1.982   0.673  0.00  0.00    +0.000 A
ATOM     30  C   UNL     1       5.077   1.126   0.476  0.00  0.00    +0.000 A
ENDBRANCH  24  25
ENDBRANCH  10  22
TORSDOF 5

I have attached files necessary to run a much faster and simpler unidock to reproduce the above error. It is worth noting that using these files also produced the constant "ENERGY FROM SEARCH", followed by high affinity scores.

files_to_reproduce.zip

Settings used

The pdbqt file: rec_1CX2_clean.pdbqt
X coordinate: 24.644557093425625
Y coordinate: 21.8624152249135
Z coordinate: 16.1463062283737
docking size X: 20
docking size Y: 20
docking size Z: 20
exhaustiveness: 512
max step: 40
number of ligand modes: 1

Expected result

[caic99]

Root Cause Analysis

The FLT_MAX energy value (3.4028235e+38) comes from out-of-bounds pose handling in the CPU refinement step.

When the GPU BFGS optimization moves a pose outside the grid box, vina.cpp runs a refinement loop with increasing penalty slopes to pull it back. If after all iterations the pose is still outside bounds, poses[i].e = max_fl is set (where fl = float, so max_fl = FLT_MAX).

The bug: these invalid poses were still being rescored and written to output. The rescoring produces the high positive affinity values (24700, 8311, etc.) from evaluating an out-of-bounds conformation.

Fix in draft PR #182: Skip poses where not_max(poses[i].e) is false at all scoring/output loops. These poses are invalid and should not appear in results.