Communications Biology,
Год журнала:
2025,
Номер
8(1)
Опубликована: Фев. 13, 2025
Abstract
The
rapid
advancement
of
single-cell
technologies
has
created
an
urgent
need
for
effective
methods
to
integrate
and
harmonize
data.
Technical
biological
variations
across
studies
complicate
data
integration,
while
conventional
tools
often
struggle
with
reliance
on
gene
expression
distribution
assumptions
over-correction.
Here,
we
present
scCobra,
a
deep
generative
neural
network
designed
overcome
these
challenges
through
contrastive
learning
domain
adaptation.
scCobra
effectively
mitigates
batch
effects,
minimizes
over-correction,
ensures
biologically
meaningful
integration
without
assuming
specific
distributions.
It
enables
online
label
transfer
datasets
allowing
continuous
new
retraining.
Additionally,
supports
effect
simulation,
advanced
multi-omic
scalable
processing
large
datasets.
By
integrating
harmonizing
from
similar
studies,
expands
the
available
investigating
problems,
improving
cross-study
comparability,
revealing
insights
that
may
be
obscured
in
isolated
Nature Biotechnology,
Год журнала:
2023,
Номер
41(12), С. 1746 - 1757
Опубликована: Март 27, 2023
Abstract
Metacells
are
cell
groupings
derived
from
single-cell
sequencing
data
that
represent
highly
granular,
distinct
states.
Here
we
present
aggregation
of
states
(SEACells),
an
algorithm
for
identifying
metacells
overcome
the
sparsity
while
retaining
heterogeneity
obscured
by
traditional
clustering.
SEACells
outperforms
existing
algorithms
in
comprehensive,
compact
and
well-separated
both
RNA
assay
transposase-accessible
chromatin
(ATAC)
modalities
across
datasets
with
discrete
types
continuous
trajectories.
We
demonstrate
use
to
improve
gene–peak
associations,
compute
ATAC
gene
scores
infer
activities
critical
regulators
during
differentiation.
Metacell-level
analysis
scales
large
is
particularly
well
suited
patient
cohorts,
where
per-patient
provides
more
robust
units
integration.
our
reveal
expression
dynamics
gradual
reconfiguration
landscape
hematopoietic
differentiation
uniquely
identify
CD4
T
activation
associated
disease
onset
severity
a
Coronavirus
Disease
2019
(COVID-19)
cohort.
bioRxiv (Cold Spring Harbor Laboratory),
Год журнала:
2023,
Номер
unknown
Опубликована: Дек. 8, 2023
Emerging
imaging
spatial
transcriptomics
(iST)
platforms
and
coupled
analytical
methods
can
recover
cell-to-cell
interactions,
groups
of
spatially
covarying
genes,
gene
signatures
associated
with
pathological
features,
are
thus
particularly
well-suited
for
applications
in
formalin
fixed
paraffin
embedded
(FFPE)
tissues.
Here,
we
benchmarked
the
performance
three
commercial
iST
on
serial
sections
from
tissue
microarrays
(TMAs)
containing
23
tumor
normal
types
both
relative
technical
biological
performance.
On
matched
found
that
10x
Xenium
shows
higher
transcript
counts
per
without
sacrificing
specificity,
but
all
concord
to
orthogonal
RNA-seq
datasets
perform
resolved
cell
typing,
albeit
different
false
discovery
rates,
segmentation
error
frequencies,
varying
degrees
sub-clustering
downstream
analyses.
Taken
together,
our
analyses
provide
a
comprehensive
benchmark
guide
choice
method
as
researchers
design
studies
precious
samples
this
rapidly
evolving
field.
