Nature Communications,
Год журнала:
2020,
Номер
11(1)
Опубликована: Фев. 5, 2020
Abstract
The
type
and
genomic
context
of
cancer
mutations
depend
on
their
causes.
These
causes
have
been
characterized
using
signatures
that
represent
mutation
types
co-occur
in
the
same
tumours.
However,
it
remains
unclear
how
processes
change
during
evolution
due
to
lack
reliable
methods
reconstruct
evolutionary
trajectories
mutational
signature
activity.
Here,
as
part
ICGC/TCGA
Pan-Cancer
Analysis
Whole
Genomes
(PCAWG)
Consortium,
which
aggregated
whole-genome
sequencing
data
from
2658
cancers
across
38
tumour
types,
we
present
TrackSig,
a
new
method
reconstructs
these
optimal,
joint
segmentation
deconvolution
allele
frequencies
single
sample.
In
simulations,
find
TrackSig
has
3–5%
activity
reconstruction
error,
12%
false
detection
rate.
It
outperforms
an
aggressive
baseline
situations
with
branching
evolution,
CNA
gain,
neutral
mutations.
Applied
tumours
permits
pan-cancer
insight
into
changes
processes.
Journal of Asthma and Allergy,
Год журнала:
2022,
Номер
Volume 15, С. 855 - 873
Опубликована: Июнь 1, 2022
Asthma
is
a
variable
long-term
condition.
Currently,
there
no
cure
for
asthma
and
the
focus
is,
therefore,
on
management.
Mobile
health
(mHealth)
promising
chronic
disease
management
but
to
be
able
realize
its
potential,
it
needs
go
beyond
simply
monitoring.
mHealth
therefore
leverage
machine
learning
provide
tailored
feedback
with
personalized
algorithms.
There
need
understand
extent
of
that
has
been
leveraged
in
context
This
review
aims
fill
this
gap.
Nature Communications,
Год журнала:
2023,
Номер
14(1)
Опубликована: Июнь 21, 2023
Abstract
Cancer
progression
is
an
evolutionary
process
shaped
by
both
deterministic
and
stochastic
forces.
Multi-region
single-cell
sequencing
of
tumors
enable
high-resolution
reconstruction
the
mutational
history
each
tumor
highlight
extensive
diversity
across
patients.
Resolving
interactions
among
mutations
recovering
recurrent
processes
may
offer
greater
opportunities
for
successful
therapeutic
strategies.
To
this
end,
we
present
a
novel
probabilistic
framework,
called
TreeMHN,
joint
inference
exclusivity
patterns
trajectories
from
cohort
intra-tumor
phylogenetic
trees.
Through
simulations,
show
that
TreeMHN
outperforms
existing
alternatives
can
only
focus
on
one
aspect
task.
By
analyzing
datasets
blood,
lung,
breast
cancers,
find
most
likely
patterns,
consistent
with
enriching
our
current
understanding
tumorigenesis.
Moreover,
facilitates
prediction
evolution
provides
measures
next
events
given
tree,
prerequisite
evolution-guided
treatment
Scientific Reports,
Год журнала:
2024,
Номер
14(1)
Опубликована: Март 10, 2024
Abstract
Cancer
presents
a
significant
global
health
burden,
resulting
in
millions
of
annual
deaths.
Timely
detection
is
critical
for
improving
survival
rates,
offering
crucial
window
timely
medical
interventions.
Liquid
biopsy,
analyzing
genetic
variations,
and
mutations
circulating
cell-free,
tumor
DNA
(cfDNA/ctDNA)
or
molecular
biomarkers,
has
emerged
as
tool
early
detection.
This
study
focuses
on
cancer
using
plasma
cfDNA/ctDNA
protein
biomarker
concentrations.
The
proposed
system
initially
calculates
the
correlation
coefficient
to
identify
correlated
features,
while
mutual
information
assesses
each
feature's
relevance
target
variable,
eliminating
redundant
features
improve
efficiency.
eXtrem
Gradient
Boosting
(XGBoost)
feature
importance
method
iteratively
selects
top
ten
60%
dataset
dimensionality
reduction.
Light
Machine
(LGBM)
model
employed
classification,
optimizing
its
performance
through
random
search
hyper-parameters.
Final
predictions
are
obtained
by
ensembling
LGBM
models
from
tenfold
cross-validation,
weighted
their
respective
balanced
accuracy,
averaged
get
final
predictions.
Applying
this
methodology,
achieves
99.45%
accuracy
99.95%
AUC
detecting
presence
achieving
93.94%
97.81%
cancer-type
classification.
Our
methodology
leads
enhanced
healthcare
outcomes
patients.
Annual Review of Genomics and Human Genetics,
Год журнала:
2019,
Номер
20(1), С. 309 - 329
Опубликована: Май 6, 2019
Cancers
originate
from
somatic
cells
in
the
human
body
that
have
accumulated
genetic
alterations.
These
mutations
modify
phenotype
of
cells,
allowing
them
to
escape
homeostatic
regulation
maintains
normal
cell
number.
Viewed
through
lens
evolutionary
biology,
transformation
into
malignant
is
evolution
action.
Evolution
continues
throughout
cancer
growth,
progression,
treatment
resistance,
and
disease
relapse,
driven
by
adaptation
changes
cancer's
environment,
intratumor
heterogeneity
an
inevitable
consequence
this
process.
Genomics
provides
a
powerful
means
characterize
tumor
evolution,
enabling
quantitative
measurement
evolving
clones
across
space
time.
In
review,
we
discuss
concepts
approaches
quantify
measure
process
using
genomics.
Extensive
transcriptional
alterations
are
observed
in
cancer,
many
of
which
activate
core
biological
processes
established
unicellular
organisms
or
suppress
differentiation
pathways
formed
metazoans.
Through
rigorous,
integrative
analysis
genomics
data
from
a
range
solid
tumors,
we
show
changes
tumors
tied
to
mutations
disrupting
regulatory
interactions
between
and
multicellular
genes
within
human
gene
networks
(GRNs).
Recurrent
point
were
enriched
regulator
linking
subnetworks,
while
copy-number
affected
downstream
target
distinctly
regions
the
GRN.
Our
results
depict
drivers
tumourigenesis
as
that
created
key
links
during
evolution
early
life,
whose
dysfunction
creates
widespread
dysregulation
primitive
elements
Several
identified
important
this
process
associated
with
drug
response,
demonstrating
potential
clinical
value
our
approach.
International Journal of Molecular Sciences,
Год журнала:
2021,
Номер
22(9), С. 4394 - 4394
Опубликована: Апрель 22, 2021
Artificial
Intelligence
is
providing
astonishing
results,
with
medicine
being
one
of
its
favourite
playgrounds.
Machine
Learning
and,
in
particular,
Deep
Neural
Networks
are
behind
this
revolution.
Among
the
most
challenging
targets
interest
cancer
diagnosis
and
therapies
but,
to
start
revolution,
software
tools
need
be
adapted
cover
new
requirements.
In
sense,
learning
becoming
a
commodity
able
assist
doctors
on
daily
basis,
it
essential
fully
understand
how
models
can
interpreted.
survey,
we
analyse
current
machine
other
in-silico
as
applied
medicine—specifically,
research—and
discuss
their
interpretability,
performance
input
data
they
fed
with.
neural
networks
(ANN),
logistic
regression
(LR)
support
vector
machines
(SVM)
have
been
observed
preferred
models.
addition,
convolutional
(CNNs),
supported
by
rapid
development
graphic
processing
units
(GPUs)
high-performance
computing
(HPC)
infrastructures,
gaining
importance
when
image
feasible.
However,
interpretability
predictions
so
that
them,
trust
them
gain
useful
insights
for
clinical
practice
still
rarely
considered,
which
factor
needs
improved
enhance
doctors’
predictive
capacity
achieve
individualised
near
future.
International Journal of Molecular Sciences,
Год журнала:
2021,
Номер
22(2), С. 924 - 924
Опубликована: Янв. 18, 2021
Diffuse
gliomas
are
the
most
frequent
brain
tumours,
representing
75%
of
all
primary
malignant
tumours
in
adults.
Because
their
locally
aggressive
behaviour
and
fact
that
they
cannot
be
cured
by
current
therapies,
represent
one
devastating
cancers.
The
present
review
summarises
recent
advances
our
understanding
glioma
development
progression
use
various
vitro
vivo
models,
as
well
more
complex
techniques
including
cultures
3D
organoids
organotypic
slices.
We
discuss
progress
has
been
made
heterogeneity,
alteration
gene
expression
DNA
methylation,
silico
models.
Lastly
treatment
options
future
clinical
trials,
which
aim
to
improve
early
diagnosis
disease
monitoring,
also
discussed.
Cells,
Год журнала:
2021,
Номер
10(4), С. 928 - 928
Опубликована: Апрель 17, 2021
Cancer
is
a
multifactorial
disease
with
increasing
incidence.
There
are
more
than
100
different
cancer
types,
defined
by
location,
cell
of
origin,
and
genomic
alterations
that
influence
oncogenesis
therapeutic
response.
This
heterogeneity
between
tumors
patients
also
the
within
same
patient's
tumor
pose
an
enormous
challenge
to
treatment.
In
this
review,
we
explore
on
longitudinal
latitudinal
axis,
reviewing
current
future
approaches
study
their
potential
support
oncologists
in
tailoring
treatment
regimen.
We
highlight
how
ideal
precision
oncology
reaching
far
beyond
knowledge
genetic
variants
inform
clinical
practice
discuss
technologies
strategies
already
available
improve
our
understanding
management
will
focus
integrating
multi-omics
suitable
vitro
models
proficiency
mimicking
endogenous
heterogeneity.