The Journal of Physical Chemistry A,
Journal Year:
2024,
Volume and Issue:
128(17), P. 3449 - 3457
Published: April 20, 2024
Machine
learning
(ML)
provides
a
great
opportunity
for
the
construction
of
models
with
improved
accuracy
in
classical
molecular
dynamics
(MD).
However,
ML
trained
model
is
limited
by
quality
and
quantity
training
data.
Generating
large
sets
accurate
ab
initio
data
can
require
significant
computational
resources.
Furthermore,
inconsistent
or
incompatible
different
accuracies
obtained
using
methods
may
lead
to
biased
unreliable
that
do
not
accurately
represent
underlying
physics.
Recently,
transfer
showed
its
potential
avoiding
these
problems
as
well
improving
accuracy,
efficiency,
generalization
multifidelity
In
this
work,
ML-based
MD
(aML-MD)
are
developed
through
DFT
multireference
from
multiple
sources
varying
within
Deep
Potential
framework.
The
force
field
demonstrated
calculating
rate
constants
H
+
HO2
→
H2
3O2
reaction
quasi-classical
trajectories.
We
show
aML-MD
predict
while
reducing
cost
more
than
five
times
compared
use
expensive
quantum
chemistry
sets.
Hence,
shows
reduce
involved
generating
set
potentials.
Chemical Science,
Journal Year:
2025,
Volume and Issue:
unknown
Published: Jan. 1, 2025
We
present
a
comprehensive
analysis
of
the
capabilities
modern
machine
learning
force
fields
to
simulate
long-term
molecular
dynamics
at
near-ambient
conditions
for
molecules,
molecule-surface
interfaces,
and
materials
within
TEA
Challenge
2023.
Nature Communications,
Journal Year:
2024,
Volume and Issue:
15(1)
Published: July 20, 2024
Abstract
Solvent
effects
influence
all
stages
of
the
chemical
processes,
modulating
stability
intermediates
and
transition
states,
as
well
altering
reaction
rates
product
ratios.
However,
accurately
modelling
these
remains
challenging.
Here,
we
present
a
general
strategy
for
generating
reactive
machine
learning
potentials
to
model
processes
in
solution.
Our
approach
combines
active
with
descriptor-based
selectors
automation,
enabling
construction
data-efficient
training
sets
that
span
relevant
conformational
space.
We
apply
this
investigate
Diels-Alder
water
methanol.
The
generated
enable
us
obtain
are
agreement
experimental
data
analyse
solvents
on
mechanism.
offers
an
efficient
routine
reactions
solution,
opening
up
avenues
studying
complex
manner.
Catalysis Science & Technology,
Journal Year:
2023,
Volume and Issue:
14(3), P. 515 - 532
Published: Dec. 1, 2023
Computational
design
of
catalytic
materials
is
a
high
dimensional
structure
optimization
problem
that
limited
by
the
bottleneck
expensive
quantum
computation
tools.
An
illustration
interaction
different
factors
involved
in
and
catalyst.
Precision Chemistry,
Journal Year:
2024,
Volume and Issue:
2(11), P. 570 - 586
Published: Sept. 11, 2024
This
Perspective
explores
the
integration
of
machine
learning
potentials
(MLPs)
in
research
heterogeneous
catalysis,
focusing
on
their
role
identifying
Journal of Chemical Theory and Computation,
Journal Year:
2023,
Volume and Issue:
20(1), P. 164 - 177
Published: Dec. 18, 2023
We
present
a
transferable
MACE
interatomic
potential
that
is
applicable
to
open-
and
closed-shell
drug-like
molecules
containing
hydrogen,
carbon,
oxygen
atoms.
Including
an
accurate
description
of
radical
species
extends
the
scope
possible
applications
bond
dissociation
energy
(BDE)
prediction,
for
example,
in
context
cytochrome
P450
(CYP)
metabolism.
The
transferability
was
validated
on
COMP6
data
set,
only
molecules,
where
it
reaches
better
accuracy
than
readily
available
general
ANI-2x
potential.
achieves
similar
two
CYP
metabolism-specific
sets,
which
include
structures.
This
model
enables
us
calculate
aliphatic
C-H
BDE,
allows
compare
reaction
energies
hydrogen
abstraction,
rate-limiting
step
hydroxylation
catalyzed
by
CYPs.
On
"CYP
3A4"
BDE
RMSE
1.37
kcal/mol
prediction
ranks
alternatives:
semiempirical
AM1
GFN2-xTB
methods
ALFABET
directly
predicts
enthalpies.
Finally,
we
highlight
smoothness
over
paths
sp
