Abstract
Optimal
integration
of
transcriptomics
data
and
associated
spatial
information
is
essential
towards
fully
exploiting
to
dissect
tissue
heterogeneity
map
out
inter-cellular
communications.
We
present
SEDR,
which
uses
a
deep
autoencoder
coupled
with
masked
self-supervised
learning
mechanism
construct
low-dimensional
latent
representation
gene
expression,
then
simultaneously
embedded
the
corresponding
through
variational
graph
autoencoder.
SEDR
achieved
higher
clustering
performance
on
manually
annotated
10
×
Visium
datasets
better
scalability
high-resolution
than
existing
methods.
Additionally,
we
show
SEDR’s
ability
impute
denoise
expression
(URL:
https://github.com/JinmiaoChenLab/SEDR/
).
Abstract
Spatial
omics
technologies
enable
a
deeper
understanding
of
cellular
organizations
and
interactions
within
tissue
interest.
These
assays
can
identify
specific
compartments
or
regions
in
with
differential
transcript
protein
abundance,
delineate
their
interactions,
complement
other
methods
defining
phenotypes.
A
variety
spatial
methodologies
are
being
developed
commercialized;
however,
these
techniques
differ
resolution,
multiplexing
capability,
scale/throughput,
coverage.
Here,
we
review
the
current
prospective
landscape
single
cell
to
subcellular
resolution
analysis
tools
provide
comprehensive
picture
for
both
research
clinical
applications.
Cell,
Год журнала:
2024,
Номер
187(8), С. 1990 - 2009.e19
Опубликована: Март 20, 2024
Multiple
sclerosis
(MS)
is
a
neurological
disease
characterized
by
multifocal
lesions
and
smoldering
pathology.
Although
single-cell
analyses
provided
insights
into
cytopathology,
evolving
cellular
processes
underlying
MS
remain
poorly
understood.
We
investigated
the
dynamics
of
modeling
temporal
regional
rates
progression
in
mouse
experimental
autoimmune
encephalomyelitis
(EAE).
By
performing
spatial
expression
profiling
using
situ
sequencing
(ISS),
we
annotated
neighborhoods
found
centrifugal
evolution
active
lesions.
demonstrated
that
disease-associated
(DA)-glia
arise
independently
are
dynamically
induced
resolved
over
course.
Single-cell
mapping
human
archival
spinal
cords
confirmed
differential
distribution
homeostatic
DA-glia,
enabled
deconvolution
inactive
sub-compartments,
identified
new
lesion
areas.
establishing
resource
neuropathology
at
resolution,
our
study
unveils
intricate
MS.
Nature Methods,
Год журнала:
2024,
Номер
21(2), С. 267 - 278
Опубликована: Янв. 8, 2024
Abstract
It
is
poorly
understood
how
different
cells
in
a
tissue
organize
themselves
to
support
functions.
We
describe
the
CytoCommunity
algorithm
for
identification
of
cellular
neighborhoods
(TCNs)
based
on
cell
phenotypes
and
their
spatial
distributions.
learns
mapping
directly
from
phenotype
space
TCN
using
graph
neural
network
model
without
intermediate
clustering
embeddings.
By
leveraging
pooling,
enables
de
novo
condition-specific
predictive
TCNs
under
supervision
sample
labels.
Using
several
types
omics
data,
we
demonstrate
that
can
identify
variable
sizes
with
substantial
improvement
over
existing
methods.
analyzing
risk-stratified
colorectal
breast
cancer
revealed
new
granulocyte-enriched
cancer-associated
fibroblast-enriched
specific
high-risk
tumors
altered
interactions
between
neoplastic
immune
or
stromal
within
TCNs.
perform
unsupervised
supervised
analyses
maps
enable
discovery
cell–cell
communication
patterns
across
scales.
Abstract
Existing
methods
for
analysis
of
spatial
transcriptomic
data
focus
on
delineating
the
global
gene
expression
variations
cell
types
across
tissue,
rather
than
local
changes
driven
by
cell-cell
interactions.
We
propose
a
new
statistical
procedure
called
niche-differential
(niche-DE)
that
identifies
cell-type-specific
niche-associated
genes,
which
are
differentially
expressed
within
specific
type
in
context
niches.
further
develop
niche-LR,
method
to
reveal
ligand-receptor
signaling
mechanisms
underlie
patterns.
Niche-DE
and
niche-LR
applicable
low-resolution
spot-based
transcriptomics
is
single-cell
or
subcellular
resolution.
Nature Communications,
Год журнала:
2024,
Номер
15(1)
Опубликована: Март 28, 2024
Abstract
Accurate
placenta
pathology
assessment
is
essential
for
managing
maternal
and
newborn
health,
but
the
placenta’s
heterogeneity
temporal
variability
pose
challenges
histology
analysis.
To
address
this
issue,
we
developed
‘Histology
Analysis
Pipeline.PY’
(HAPPY),
a
deep
learning
hierarchical
method
quantifying
of
cells
micro-anatomical
tissue
structures
across
whole
slide
images.
HAPPY
differs
from
patch-based
features
or
segmentation
approaches
by
following
an
interpretable
biological
hierarchy,
representing
cellular
communities
within
tissues
at
single-cell
resolution
We
present
set
quantitative
metrics
healthy
term
placentas
as
baseline
future
assessments
health
show
how
these
deviate
in
with
clinically
significant
placental
infarction.
HAPPY’s
cell
predictions
closely
replicate
those
independent
clinical
experts
biology
literature.
