Development and external validation of a machine learning-based model to predict postoperative recurrence in patients with duodenal adenocarcinoma: a multicenter, retrospective cohort study
BMC Medicine,
Год журнала:
2025,
Номер
23(1)
Опубликована: Фев. 21, 2025
Duodenal
adenocarcinoma
(DA)
has
a
high
recurrence
rate,
making
the
prediction
of
after
surgery
critically
important.
Our
objective
is
to
develop
machine
learning-based
model
predict
postoperative
DA.
We
conducted
multicenter,
retrospective
cohort
study
in
China.
1830
patients
with
DA
who
underwent
radical
between
2012
and
2023
were
included.
Wrapper
methods
used
select
optimal
predictors
by
ten
learning
learners.
Subsequently,
these
learners
utilized
for
development.
The
model's
performance
was
validated
using
three
separate
cohorts,
assessed
concordance
index
(C-index),
time-dependent
calibration
curve,
receiver
operating
characteristic
curves,
decision
curve
analysis.
After
selecting
predictors,
feature
subsets
identified.
And
combined
permutation,
resulting
development
100
predictive
models,
Penalized
Regression
+
Accelerated
Oblique
Random
Survival
Forest
(PAM)
exhibited
best
performance.
C-index
PAM
0.882
(95%
CI
0.860–0.886)
training
cohort,
0.747
0.683–0.798)
validation
1,
0.736
0.649–0.792)
2,
0.734
0.674–0.791)
3.
A
publicly
accessible
web
tool
developed
PAM.
potential
identify
patients.
This
can
assist
clinicians
assessing
severity
disease,
facilitating
patient
follow-up,
aiding
formulation
adjuvant
treatment
strategies.
Язык: Английский
Potential role of CFLAR in enhancing 5-FU sensitivity and modulating immune cell infiltration in breast cancer
European journal of medical research,
Год журнала:
2025,
Номер
30(1)
Опубликована: Апрель 10, 2025
Breast
cancer
(BRCA),
the
most
common
malignancy
among
women,
is
a
highly
heterogeneous
disease.
Chemoresistance
major
factor
leading
to
treatment
failure
in
BRCA.
However,
mechanisms
underlying
development
of
chemoresistance
remain
unclear.
In
this
study,
we
performed
comprehensive
bioinformatic
analysis
examine
role
cell
death-associated
genes
BRCA
treatment.
Specifically,
focused
on
caspase
8
and
Fas-associated
protein
with
death
domain-like
apoptosis
regulator
(CFLAR),
which
was
identified
as
co-differentially
expressed
molecule
potential
prognostic
values.
We
then
validated
these
findings
through
vitro
experiments
BT-
549
MDA-MB-
231
breast
cells.
Based
bioinformatics
analysis,
CFLAR
expression
found
be
downregulated
patients
BRCA,
whereas
its
high
significantly
associated
improved
prognosis.
Kyoto
Encyclopedia
Genes
Genomes
(KEGG)
pathway
indicated
that
aberrantly
potentially
oxidative
phosphorylation,
T
receptor
signaling,
NADH
dehydrogenase
(ubiquinone)
activity.
demonstrated
overexpression
inhibited
generation
reactive
oxygen
species
(ROS),
consequently
promoting
5-fluorouracil
(5-FU)
sensitivity
The
positively
correlated
abundance
several
tumor-infiltrating
immune
cells,
especially
CD8
+
further
supporting
regulation.
conclusion,
study
reveals
novel
roles
enhancing
chemotherapy
patient
outcome
underscores
therapeutic
target.
These
results
supported
target
biomarker
patients.
Язык: Английский