Computational Biology and Chemistry,
Journal Year:
2024,
Volume and Issue:
109, P. 108023 - 108023
Published: Feb. 7, 2024
AI-enhanced
bioinformatics
and
cheminformatics
pivots
on
generating
increasingly
descriptive
generalized
molecular
representation.
Accurate
prediction
of
properties
needs
a
comprehensive
description
geometry.
We
design
novel
Graph
Isomorphic
Network
(GIN)
based
model
integrating
three-level
network
structure
with
dual-level
pre-training
approach
that
aligns
the
characteristics
molecules.
In
our
Spatial
Molecular
Pre-training
(SMPT)
Model,
can
learn
implicit
geometric
information
in
layers
from
lower
to
higher
according
dimension.
Extensive
evaluations
against
established
baseline
models
validate
enhanced
efficacy
SMPT,
notable
accomplishments
classification
tasks.
These
results
emphasize
importance
spatial
representation
modeling
demonstrate
potential
SMPT
as
valuable
tool
for
property
prediction.
Biomaterials Advances,
Journal Year:
2023,
Volume and Issue:
151, P. 213428 - 213428
Published: April 24, 2023
More
than
fifty
years
after
the
3Rs
definition
and
despite
continuous
implementation
of
regulatory
measures,
animals
continue
to
be
widely
used
in
basic
research.
Their
use
comprises
not
only
vivo
experiments
with
animal
models,
but
also
production
a
variety
supplements
products
origin
for
cell
tissue
culture,
cell-based
assays,
therapeutics.
The
animal-derived
most
research
are
fetal
bovine
serum
(FBS),
extracellular
matrix
proteins
such
as
Matrigel™,
antibodies.
However,
their
raises
several
ethical
issues
regarding
welfare.
Additionally,
biological
is
associated
high
risk
contamination,
resulting,
frequently,
poor
scientific
data
clinical
translation.
These
support
search
new
animal-free
able
replace
FBS,
antibodies
In
addition,
silico
methodologies
play
an
important
role
reduction
by
refining
previously
vitro
experiments.
this
review,
we
depicted
current
available
alternatives
Environmental Science & Technology,
Journal Year:
2023,
Volume and Issue:
57(46), P. 18067 - 18079
Published: June 6, 2023
Nontarget
high-resolution
mass
spectrometry
screening
(NTS
HRMS/MS)
can
detect
thousands
of
organic
substances
in
environmental
samples.
However,
new
strategies
are
needed
to
focus
time-intensive
identification
efforts
on
features
with
the
highest
potential
cause
adverse
effects
instead
most
abundant
ones.
To
address
this
challenge,
we
developed
MLinvitroTox,
a
machine
learning
framework
that
uses
molecular
fingerprints
derived
from
fragmentation
spectra
(MS2)
for
rapid
classification
unidentified
HRMS/MS
as
toxic/nontoxic
based
nearly
400
target-specific
and
over
100
cytotoxic
endpoints
ToxCast/Tox21.
Model
development
results
demonstrated
using
customized
models,
quarter
toxic
majority
associated
mechanistic
targets
could
be
accurately
predicted
sensitivities
exceeding
0.95.
Notably,
SIRIUS
xboost
(Extreme
Gradient
Boosting)
models
SMOTE
(Synthetic
Minority
Oversampling
Technique)
handling
data
imbalance
were
universally
successful
robust
modeling
configuration.
Validation
MLinvitroTox
MassBank
showed
toxicity
MS2
an
average
balanced
accuracy
0.75.
By
applying
data,
confirmed
experimental
obtained
target
analysis
narrowed
analytical
tens
detected
signals
783
linked
toxicity,
including
109
spectral
matches
30
compounds
activity.
Toxicological Sciences,
Journal Year:
2024,
Volume and Issue:
200(2), P. 228 - 234
Published: May 7, 2024
Abstract
Arguably
the
most
famous
principle
of
toxicology
is
“The
dose
makes
poison”
formulated
by
Paracelsus
in
16th
century.
Application
Paracelsus’s
to
mechanistic
may
be
challenging
as
one
compound
affect
many
molecular
pathways
at
different
doses
with
and
often
nonlinear
dose-response
relationships.
As
a
result,
studies
environmental
occupational
compounds
use
high
xenobiotics
motivated
need
see
clear
signal
indicating
disruption
particular
pathway.
This
approach
ignores
possibility
that
same
xenobiotic
mechanism(s)
much
lower
relevant
human
exposures.
To
amend
simple
concise
guiding
principle,
I
suggest
recontextualization
following
its
letter
spirit:
disrupts
pathway”.
Justification
this
statement
includes
observations
broad
range
cascades,
are
sensitive
chemical
exposures,
compound.
become
useful
guidance
educational
tool
toxicological
applications,
including
experimental
design,
comparative
analysis
hypotheses,
evaluation
quality
studies,
risk
assessment.
Chemical Research in Toxicology,
Journal Year:
2023,
Volume and Issue:
36(6), P. 838 - 847
Published: April 24, 2023
An
adverse
outcome
pathway
(AOP)
framework
can
be
applied
as
an
efficient
tool
for
the
rapid
screening
of
environmental
chemicals.
For
development
AOP,
a
database
mining
approach
support
expert
derivation
by
gathering
wider
range
evidence
than
in
literature
review.
In
this
study,
data
from
various
databases
were
integrated
and
analyzed
to
supplement
AOP
leading
pulmonary
fibrosis
analyzing
additional
using
establishing
application
domain
First,
we
collected
chemicals,
genes,
phenotypes
that
studied
related
through
Comparative
Toxicogenomics
Database
(CTD).
CGPD-tetramers
constructed
linking
each
chemical,
gene,
phenotype,
disease
provide
basic
components
assembly
putative
AOPs.
Next,
network
was
established
connecting
eight
existing
AOPs
developed
Wiki.
Finally,
proposed
integrating
Wiki
CTD.
To
prioritize
potential
chemical
stressors
network,
61
chemicals
ranked
relevance
exposure
information
CompTox
Chemicals
Dashboard.
The
study
guide
utilization
available
well
constructing
networks
specific
diseases.
