Chemical Science,
Journal Year:
2024,
Volume and Issue:
unknown
Published: Jan. 1, 2024
A
data-science-centered
“design–discover–evaluate”
scheme
is
presented,
and
9
novel
polyimides
suitable
for
application
to
high-temperature
energy
storage
dielectrics
are
identified
from
the
designed
virtual
structure
library.
Science,
Journal Year:
2024,
Volume and Issue:
384(6693)
Published: March 7, 2024
Deep-learning
methods
have
revolutionized
protein
structure
prediction
and
design
but
are
presently
limited
to
protein-only
systems.
We
describe
RoseTTAFold
All-Atom
(RFAA),
which
combines
a
residue-based
representation
of
amino
acids
DNA
bases
with
an
atomic
all
other
groups
model
assemblies
that
contain
proteins,
nucleic
acids,
small
molecules,
metals,
covalent
modifications,
given
their
sequences
chemical
structures.
By
fine-tuning
on
denoising
tasks,
we
developed
RFdiffusion
(RFdiffusionAA),
builds
structures
around
molecules.
Starting
from
random
distributions
acid
residues
surrounding
target
designed
experimentally
validated,
through
crystallography
binding
measurements,
proteins
bind
the
cardiac
disease
therapeutic
digoxigenin,
enzymatic
cofactor
heme,
light-harvesting
molecule
bilin.
iScience,
Journal Year:
2022,
Volume and Issue:
26(1), P. 105920 - 105920
Published: Dec. 30, 2022
A
crucial
component
in
structure-based
drug
discovery
is
the
availability
of
high-quality
three-dimensional
structures
protein
target.
Whenever
experimental
were
not
available,
homology
modeling
has
been,
so
far,
method
choice.
Recently,
AlphaFold
(AF),
an
artificial-intelligence-based
structure
prediction
method,
shown
impressive
results
terms
model
accuracy.
This
outstanding
success
prompted
us
to
evaluate
how
accurate
AF
models
are
from
perspective
docking-based
discovery.
We
compared
high-throughput
docking
(HTD)
performance
with
their
corresponding
PDB
using
a
benchmark
set
22
targets.
The
showed
consistently
worse
four
programs
and
two
consensus
techniques.
Although
shows
remarkable
ability
predict
architecture,
this
might
be
enough
guarantee
that
can
reliably
used
for
HTD,
post-modeling
refinement
strategies
key
increase
chances
success.
bioRxiv (Cold Spring Harbor Laboratory),
Journal Year:
2023,
Volume and Issue:
unknown
Published: Oct. 9, 2023
Abstract
Although
AlphaFold2
(AF2)
and
RoseTTAFold
(RF)
have
transformed
structural
biology
by
enabling
high-accuracy
protein
structure
modeling,
they
are
unable
to
model
covalent
modifications
or
interactions
with
small
molecules
other
non-protein
that
can
play
key
roles
in
biological
function.
Here,
we
describe
All-Atom
(RFAA),
a
deep
network
capable
of
modeling
full
assemblies
containing
proteins,
nucleic
acids,
molecules,
metals,
given
the
sequences
polymers
atomic
bonded
geometry
modifications.
Following
training
on
structures
Protein
Data
Bank
(PDB),
RFAA
has
comparable
prediction
accuracy
AF2,
excellent
performance
CAMEO
for
flexible
backbone
molecule
docking,
reasonable
proteins
multiple
acid
chains
which,
our
knowledge,
no
existing
method
simultaneously.
By
fine-tuning
diffusive
denoising
tasks,
develop
RFdiffusion
(RFdiffusionAA
)
,
which
generates
binding
pockets
directly
building
around
molecules.
Starting
from
random
distributions
amino
residues
surrounding
target
design
experimentally
validate
bind
cardiac
disease
therapeutic
digoxigenin,
enzymatic
cofactor
heme,
optically
active
bilin
potential
expanding
range
wavelengths
captured
photosynthesis.
We
anticipate
RFdiffusionAA
will
be
widely
useful
designing
complex
biomolecular
systems.
Science,
Journal Year:
2024,
Volume and Issue:
384(6702)
Published: May 16, 2024
AlphaFold2
(AF2)
models
have
had
wide
impact
but
mixed
success
in
retrospective
ligand
recognition.
We
prospectively
docked
large
libraries
against
unrefined
AF2
of
the
σ
Computational
prediction
of
protein
structure
has
been
pursued
intensely
for
decades,
motivated
largely
by
the
goal
using
structural
models
drug
discovery.
Recently
developed
machine-learning
methods
such
as
AlphaFold
2
(AF2)
have
dramatically
improved
prediction,
with
reported
accuracy
approaching
that
experimentally
determined
structures.
To
what
extent
do
these
advances
translate
to
an
ability
predict
more
accurately
how
drugs
and
candidates
bind
their
target
proteins?
Here,
we
carefully
examine
utility
AF2
predicting
binding
poses
drug-like
molecules
at
largest
class
targets,
G-protein-coupled
receptors.
We
find
capture
pocket
structures
much
than
traditional
homology
models,
errors
nearly
small
differences
between
same
different
ligands
bound.
Strikingly,
however,
ligand-binding
predicted
computational
docking
is
not
significantly
higher
when
lower
without
These
results
important
implications
all
those
who
might
use
Nature Machine Intelligence,
Journal Year:
2024,
Volume and Issue:
6(5), P. 558 - 567
Published: May 8, 2024
Abstract
Advances
in
deep
learning
have
greatly
improved
structure
prediction
of
molecules.
However,
many
macroscopic
observations
that
are
important
for
real-world
applications
not
functions
a
single
molecular
but
rather
determined
from
the
equilibrium
distribution
structures.
Conventional
methods
obtaining
these
distributions,
such
as
dynamics
simulation,
computationally
expensive
and
often
intractable.
Here
we
introduce
framework,
called
Distributional
Graphormer
(DiG),
an
attempt
to
predict
systems.
Inspired
by
annealing
process
thermodynamics,
DiG
uses
neural
networks
transform
simple
towards
distribution,
conditioned
on
descriptor
system
chemical
graph
or
protein
sequence.
This
framework
enables
efficient
generation
diverse
conformations
provides
estimations
state
densities,
orders
magnitude
faster
than
conventional
methods.
We
demonstrate
several
tasks,
including
conformation
sampling,
ligand
catalyst–adsorbate
sampling
property-guided
generation.
presents
substantial
advancement
methodology
statistically
understanding
systems,
opening
up
new
research
opportunities
sciences.
Computational
prediction
of
protein
structure
has
been
pursued
intensely
for
decades,
motivated
largely
by
the
goal
using
structural
models
drug
discovery.
Recently
developed
machine-learning
methods
such
as
AlphaFold
2
(AF2)
have
dramatically
improved
prediction,
with
reported
accuracy
approaching
that
experimentally
determined
structures.
To
what
extent
do
these
advances
translate
to
an
ability
predict
more
accurately
how
drugs
and
candidates
bind
their
target
proteins?
Here,
we
carefully
examine
utility
AF2
predicting
binding
poses
drug-like
molecules
at
largest
class
targets,
G-protein-coupled
receptors.
We
find
capture
pocket
structures
much
than
traditional
homology
models,
errors
nearly
small
differences
between
same
different
ligands
bound.
Strikingly,
however,
ligand-binding
predicted
computational
docking
is
not
significantly
higher
when
lower
without
These
results
important
implications
all
those
who
might
use
Current Opinion in Structural Biology,
Journal Year:
2024,
Volume and Issue:
85, P. 102776 - 102776
Published: Feb. 8, 2024
The
complex
molecular
mechanism
and
pathophysiology
of
Alzheimer's
disease
(AD)
limits
the
development
effective
therapeutics
or
prevention
strategies.
Artificial
Intelligence
(AI)-guided
drug
discovery
combined
with
genetics/multi-omics
(genomics,
epigenomics,
transcriptomics,
proteomics,
metabolomics)
analysis
contributes
to
understanding
precision
medicine
disease,
including
AD
AD-related
dementia.
In
this
review,
we
summarize
AI-driven
methodologies
for
AD-agnostic
development,
de
novo
design,
virtual
screening,
prediction
drug-target
interactions,
all
which
have
shown
potentials.
particular,
AI-based
repurposing
emerges
as
a
compelling
strategy
identify
new
indications
existing
drugs
AD.
We
provide
several
emerging
targets
from
human
genetics
multi-omics
findings
highlight
recent
technologies
their
applications
in
using
prototypical
example.
closing,
discuss
future
challenges
directions
other
neurodegenerative
diseases.
Science Advances,
Journal Year:
2024,
Volume and Issue:
10(32)
Published: Aug. 7, 2024
Artificial
intelligence
is
revolutionizing
protein
structure
prediction,
providing
unprecedented
opportunities
for
drug
design.
To
assess
the
potential
impact
on
ligand
discovery,
we
compared
virtual
screens
using
structures
generated
by
AlphaFold
machine
learning
method
and
traditional
homology
modeling.
More
than
16
million
compounds
were
docked
to
models
of
trace
amine-associated
receptor
1
(TAAR1),
a
G
protein-coupled
unknown
target
treating
neuropsychiatric
disorders.
Sets
30
32
highly
ranked
from
model
screens,
respectively,
experimentally
evaluated.
Of
these,
25
TAAR1
agonists
with
potencies
ranging
12
0.03
μM.
The
screen
yielded
more
twofold
higher
hit
rate
(60%)
discovered
most
potent
agonists.
A
agonist
promising
selectivity
profile
drug-like
properties
showed
physiological
antipsychotic-like
effects
in
wild-type
but
not
knockout
mice.
These
results
demonstrate
that
can
accelerate
discovery.