The Computer Journal,
Journal Year:
2021,
Volume and Issue:
66(1), P. 245 - 266
Published: Sept. 21, 2021
Abstract
This
paper
devises
a
novel
technique,
namely
Squirrel
Search
Deer
Hunting-based
deep
recurrent
neural
network
(SSDH-based
DRNN)
for
cancer-survival
rate
prediction
using
gene
expression
(GE)
data.
Initially,
the
input
GE
data
are
transformed
polynomial
kernel
transformation.
Then
entropy-based
Bayesian
fuzzy
clustering
is
employed
selection.
Then,
selected
features
strengthened
through
survival
indicators
based
on
time
series
features,
like
simple
moving
average
(SMA)
and
of
change.
Finally,
performed
(DRNN),
in
which
training
carried
out
with
squirrel
search
deer
hunting
(SSDH).
The
proposed
SSDH
algorithm
devised
by
combining
Algorithm
(SSA)
optimization
(DHOA).
performance
methodology
analyzed
Pan-Cancer
(PANCAN)
dataset
error
4.05%,
RMSE
7.58,
accuracy
90.98%,
precision
90.80%,
recall
92.03%
F1-score
91.41%.
method
higher
lower
cancer
patients
prognosis.
Besides,
it
will
be
helpful
clinical
management
patients.
Computational and Structural Biotechnology Journal,
Journal Year:
2021,
Volume and Issue:
19, P. 3735 - 3746
Published: Jan. 1, 2021
Increased
availability
of
high-throughput
technologies
has
generated
an
ever-growing
number
omics
data
that
seek
to
portray
many
different
but
complementary
biological
layers
including
genomics,
epigenomics,
transcriptomics,
proteomics,
and
metabolomics.
New
insight
from
these
have
been
obtained
by
machine
learning
algorithms
produced
diagnostic
classification
biomarkers.
Most
biomarkers
date
however
only
include
one
omic
measurement
at
a
time
thus
do
not
take
full
advantage
recent
multi-omics
experiments
now
capture
the
entire
complexity
systems.
Multi-omics
integration
strategies
are
needed
combine
knowledge
brought
each
layer.
We
summarized
most
methods/
frameworks
into
five
strategies:
early,
mixed,
intermediate,
late
hierarchical.
In
this
mini-review,
we
focus
on
challenges
existing
paying
special
attention
applications.
IEEE Reviews in Biomedical Engineering,
Journal Year:
2023,
Volume and Issue:
17, P. 80 - 97
Published: Oct. 12, 2023
With
the
recent
advancement
of
novel
biomedical
technologies
such
as
high-throughput
sequencing
and
wearable
devices,
multi-modal
data
ranging
from
multi-omics
molecular
to
real-time
continuous
bio-signals
are
generated
at
an
unprecedented
speed
scale
every
day.
For
first
time,
these
able
make
precision
medicine
close
a
reality.
However,
due
volume
complexity,
making
good
use
requires
major
effort.
Researchers
clinicians
actively
developing
artificial
intelligence
(AI)
approaches
for
data-driven
knowledge
discovery
causal
inference
using
variety
modalities.
These
AI-based
have
demonstrated
promising
results
in
various
healthcare
applications.
In
this
review
paper,
we
summarize
state-of-the-art
AI
models
integrating
electronic
health
records
(EHRs)
medicine.
We
discuss
challenges
opportunities
with
EHRs
future
directions.
hope
can
inspire
research
BMC Genomics,
Journal Year:
2022,
Volume and Issue:
23(1)
Published: April 9, 2022
Disclosure
of
patients'
genetic
information
in
the
process
applying
machine
learning
techniques
for
tumor
classification
hinders
privacy
personal
information.
Homomorphic
Encryption
(HE),
which
supports
operations
between
encrypted
data,
can
be
used
as
one
tools
to
perform
such
computation
without
leakage,
but
it
brings
great
challenges
directly
general
algorithms
due
limitations
supported
by
HE.
In
particular,
non-polynomial
activation
functions,
including
softmax
are
difficult
implement
with
HE
and
require
a
suitable
approximation
method
minimize
loss
accuracy.
secure
genome
analysis
competition
called
iDASH
2020,
is
presented
task
that
multi-label
predicts
class
samples
based
on
using
HE.We
develop
ensure
during
all
computations
model
inference
process.
Our
solution
1-layer
neural
network
function
uses
approximate
scheme.
We
present
an
enables
technique
efficiently
encoding
data
reduce
computational
costs.
addition,
we
propose
HE-friendly
filtering
size
large-scale
data.We
aim
analyze
dataset
from
The
Cancer
Genome
Atlas
(TCGA)
dataset,
consists
3,622
11
types
cancers,
features
25,128
genes.
preprocessing
reduces
number
genes
4,096
or
less
achieves
microAUC
value
0.9882
(85%
accuracy)
shallow
network.
Using
our
model,
successfully
compute
steps
test
3.75
minutes.
As
result
exceptionally
high
values,
was
awarded
co-first
place
2020
Track
1:
"Secure
Tumor
Encryption".Our
first
implementing
Also,
optimization
methods
this
work
enable
implementation
other
challenging
applications.
Applied Sciences,
Journal Year:
2024,
Volume and Issue:
14(13), P. 5660 - 5660
Published: June 28, 2024
Cancer
research
has
increasingly
utilized
multi-omics
analysis
in
recent
decades
to
obtain
biomolecular
information
from
multiple
layers,
thereby
gaining
a
better
understanding
of
complex
biological
systems.
However,
the
curse
dimensionality
is
one
most
significant
challenges
when
handling
omics
or
data.
Additionally,
integrating
by
transforming
different
types
into
new
representation
can
reduce
model’s
interpretability,
as
extracted
features
may
lose
context.
This
paper
proposes
Iterative
Similarity
Bagging
(ISB),
assisted
Bayesian
Ridge
Regression
(BRR).
BRR
serves
domain-oriented
supervised
feature
selection
method,
choosing
essential
calculating
coefficients
for
each
feature.
Despite
this,
output
datasets
contain
many
features,
leading
complexity
and
high
dimensionality.
To
address
ISB
was
introduced
dynamically
without
losing
integrity
data,
which
often
occurs
with
transformation-based
integration
approaches.
The
evaluation
measures
employed
were
Root
Mean
Square
Error
(RMSE),
Pearson
Correlation
Coefficient
(PCC),
coefficient
determination
(R2).
results
demonstrate
that
proposed
method
outperforms
some
current
models
terms
regression
performance,
achieving
an
RMSE
0.12,
PCC
0.879,
R2
0.77
CCLE.
For
GDSC,
it
achieved
0.029,
0.90,
0.80.
Frontiers in Medicine,
Journal Year:
2023,
Volume and Issue:
9
Published: Jan. 10, 2023
Gastrointestinal
cancer
is
becoming
increasingly
common,
which
leads
to
over
3
million
deaths
every
year.
No
typical
symptoms
appear
in
the
early
stage
of
gastrointestinal
cancer,
posing
a
significant
challenge
diagnosis
and
treatment
patients
with
cancer.
Many
are
middle
late
stages
when
they
feel
uncomfortable,
unfortunately,
most
them
will
die
Recently,
various
artificial
intelligence
techniques
like
machine
learning
based
on
multi-omics
have
been
presented
for
era
precision
medicine.
This
paper
provides
survey
multi-omics-based
using
potential
application
Particularly,
we
make
comprehensive
summary
analysis
from
perspective
datasets,
task
types,
integration
methods.
Furthermore,
this
points
out
remaining
challenges
discusses
future
topics.
Computers in Biology and Medicine,
Journal Year:
2024,
Volume and Issue:
179, P. 108818 - 108818
Published: July 10, 2024
Breast
cancer
is
the
most
common
malignant
neoplasm
and
leading
cause
of
mortality
among
women
globally.
Current
prediction
models
based
on
risk
factors
are
inefficient
in
specific
populations,
so
an
appropriate
calibrated
breast
model
for
Cuban
essential.
This
article
proposes
a
conceptual
estimation
using
machine
learning
algorithms
factors.
The
has
three
main
components:
knowledge
representation,
modeling,
predictor
evaluation.
Nine
were
used
to
generate
predictors
proposed
model.
Two
data
sources
served
as
case
studies:
first
comprised
collected
from
women,
second
included
US
Hispanic
obtained
Cancer
Surveillance
Consortium
dataset.
results
show
that
effectively
estimates
could
be
valuable
tool
early
detection
identification
patients
at
risk.
According
experiment
results,
best
female
population
corresponds
Random
Forest
algorithm
with
weighted
score
5.981,
training
accuracy
0.996
AUC
0.997.
In
experiment,
it
was
demonstrated
generated
by
better
values
compared
population,
potentially
generalizable
other
populations.
Implementing
this
economically
viable
alternative
reduce
rate
type
Latin
American
countries
such
Cuba.
2021 IEEE International Symposium on Software Reliability Engineering Workshops (ISSREW),
Journal Year:
2022,
Volume and Issue:
unknown, P. 289 - 294
Published: Oct. 1, 2022
Detecting
the
gene
sequence
of
virus
strains
from
patients
and
classifying
them
into
specific
are
very
important
to
provide
effective
treatment.
However,
there
significant
barriers
sharing
strains'
data
in
plaintext
privacy
concerns
patients.
Homomorphic
encryption
is
a
form
that
allows
users
calculate
encrypted
without
decrypting
it.
Achieving
highly
accurate
viral
strain
prediction
while
safeguarding
user
challenge.
We
develop
secure
multi-label
classification
method
using
homomorphic
scheme.
first
used
statistical
genotype
frequencies
for
preprocessing
reduce
dimension
strains.
Second,
we
improved
TFHE
library
proposed
by
Chillotti
et
al.
accommodate
floating-point
input
neural
network
make
calculation
result
more
accurate.
Finally,
improve
computational
speed
storage
usage
packing
packs
multiple
feature
information
one
ciphertext.
successfully
calculated
2000
inference
steps
on
128-bit
test
0.09
seconds,
reaching
an
accuracy
100
%.