The
performance
of
electrochemical
sensors
is
influenced
by
various
factors.
To
enhance
the
effectiveness
these
sensors,
it
crucial
to
find
right
balance
among
Researchers
and
engineers
continually
explore
innovative
approaches
sensitivity,
selectivity,
reliability.
Machine
learning
(ML)
techniques
facilitate
analysis
predictive
modeling
sensor
establishing
quantitative
relationships
between
parameters
their
effects.
This
work
presents
a
case
study
on
developing
molecularly
imprinted
polymer
(MIP)-based
for
detecting
doxorubicin
(Dox),
emphasizing
use
ML-based
ensemble
models
improve
Four
ML
models,
including
Decision
Tree
(DT),
eXtreme
Gradient
Boosting
(XGBoost),
Random
Forest
(RF),
K-Nearest
Neighbors
(KNN),
are
used
evaluate
effect
each
parameter
prediction
performance,
using
SHapley
Additive
exPlanations
(SHAP)
method
determine
feature
importance.
Based
analysis,
removing
less
influential
introducing
new
significantly
improved
model's
capabilities.
By
applying
min-max
scaling
technique,
ensured
that
all
features
contribute
proportionally
model
process.
Additionally,
multiple
models─Linear
Regression
(LR),
KNN,
DT,
RF,
Adaptive
(AdaBoost),
(GB),
Support
Vector
(SVR),
XGBoost,
Bagging,
Partial
Least
Squares
(PLS),
Ridge
Regression─are
applied
data
set
in
predicting
output
current
compared.
further
novel
proposed
integrates
GB,
Bagging
regressors,
leveraging
combined
strengths
offset
individual
weaknesses.
main
benefit
this
lies
its
ability
MIP-based
stacking
regressor
model,
which
improves
methodology
broadly
applicable
development
other
with
different
transducers
sensing
elements.
Through
extensive
simulation
results,
demonstrated
superior
compared
models.
achieved
an
R-squared
(R2)
0.993,
reducing
root-mean-square
error
(RMSE)
0.436
mean
absolute
(MAE)
0.244.
These
improvements
enhanced
sensitivity
reliability
sensor,
demonstrating
substantial
gain
over
Gels,
Год журнала:
2025,
Номер
11(2), С. 114 - 114
Опубликована: Фев. 6, 2025
Silk
sericin
(SS),
a
by-product
of
the
textile
industry,
has
gained
significant
attention
for
its
biomedical
potential
due
to
biocompatibility
and
regenerative
potential.
However,
literature
lacks
information
on
SS
processing
methods
resulting
physicochemical
properties.
This
study
represents
first
step
in
protocol
optimization
standardization.
In
present
work,
different
techniques
were
studied
compared
extracted
from
boiling
water:
evaporation,
rotary
lyophilization,
dialysis,
which
presented
recovery
yield
approximately
27–32%.
The
goal
was
find
most
promising
process
concentrate
solutions,
ensure
that
structure
highly
preserved.
As
result,
new
cryo-lyophilization
methodology
proposed.
proposed
method
allows
preservation
amorphous
structure,
offers
advantages
including
complete
dissolution
water
PBS,
an
increase
storage
stability,
possibility
scaling-up,
making
it
suitable
industrial
applications.
second
part
work
focused
addressing
another
challenge
processing:
efficient
non-destructive
sterilization.
Supercritical
CO2
(scCO2)
been
gaining
momentum
last
years
sterilizing
sensitive
biopolymers
biological
materials
non-toxicity
mild
conditions.
Thus,
scCO2
technology
validated
as
technique
terminal
sterilization
SS.
this
way,
possible
engineer
sequential
cryo-lyophilization/scCO2
able
preserve
original
properties
natural
silk
protein.
Overall,
we
have
valorized
into
sterile,
off-the-shelf,
bioactive,
water-soluble
material,
with
be
used
biomedical,
pharmaceutical,
or
cosmetic
industries.
Nanotechnology
enables
targeted
theranostics
by
enhancing
imaging
and
therapy.
This
review
explores
advances
in
pH-induced
charge-switchable
nanomaterials,
their
role
drug
delivery
contrast,
future
directions
the
field.
PeerJ Computer Science,
Год журнала:
2025,
Номер
11, С. e2742 - e2742
Опубликована: Март 3, 2025
Sports
behavior
prediction
requires
precise
and
reliable
analysis
of
muscle
activity
during
exercise.
This
study
proposes
a
multi-channel
correlation
feature
extraction
method
for
electromyographic
(EMG)
signals
to
overcome
challenges
in
sports
prediction.
A
wavelet
threshold
denoising
algorithm
is
enhanced
with
nonlinear
function
transitions
control
coefficients
improve
signal
quality,
achieving
effective
noise
reduction
higher
signal-to-noise
ratio.
Furthermore,
linear
features
are
combined,
leveraging
mutual
information
estimation
via
copula
entropy
construction.
stacking
ensemble
learning
model,
incorporating
extreme
gradient
boosting
(XGBoost),
K-nearest
network
(KNN),
Random
Forest
(RF),
naive
Bayes
(NB)
as
base
learners,
further
enhances
classification
accuracy.
Experimental
results
demonstrate
that
the
proposed
approach
achieves
over
95%
accuracy,
significantly
outperforming
traditional
methods.
The
robustness
validated
across
diverse
datasets,
proving
their
effectiveness
mitigating
channel
crosstalk
interference.
work
provides
scientific
basis
improving
training
strategies
reducing
injury
risks.
