Table 1 Computational chemistry methods in our quantum refinementsa

From: Accelerating reliable multiscale quantum refinement of protein–drug systems enabled by machine learning

 

Methods

M1

ONIOM2(DFT:MM)

M2

ONIOM2-EE(DFT:MM)

M3

ONIOM2(ANI-2x:MM)

M4

ONIOM2(ANI-1ccx:MM)

M4a

ONIOM3(ANI-1ccx:ANI-2x:MM)b

M5

ONIOM2(AIQM1:MM)c

M5a

ONIOM3(AIQM1:ANI-2x:MM)b

M6

ONIOM2(SE:MM)

M7

ONIOM3(DFT:SE:MM)

M8

ONIOM3(ANI-2x:SE:MM)

M9

ONIOM3(ANI-1ccx:SE:MM)

M9a

ONIOM4(ANI-1ccx:ANI-2x:SE:MM)b

M10

ONIOM3(AIQM1:SE:MM)c

M10a

ONIOM4(AIQM1:ANI-2x:SE:MM)b

M6R

ONIOM2(SE:MM)d

  1. aωB97X-D/6-31 G(d) as the density functional theory (DFT) method, ANI-2x (machine learning potentials with DFT accuracy, MLP-DFT), ANI-1ccx (machine learning potentials with coupled-cluster accuracy, MLP-CC) and AIQM1 (MLP-CC) as machine learning potentials (MLPs), Amber ff14SB as the MM method, and GFN2-xTB as the semi-empirical (SE) method.
  2. bTwo machine learning potentials were used to describe the drug/inhibitor molecules due to the element limitations (C, H, O, N) of ANI-1ccx and AIQM1.
  3. cAIQM1 combined with GFN2-xTB methods were used to describe the drug/inhibitor molecules containing P or Br elements due to the element limitations (C, H, O, N) of AIQM1.
  4. dThe high layer of M6R includes the drug/inhibitor molecule and its neighboring residues within 3 Å.
Back to article page