Vasculature segmentation in 3D hierarchical phase-contrast tomography images of human kidneys
bioRxiv (Cold Spring Harbor Laboratory),
Год журнала:
2024,
Номер
unknown
Опубликована: Авг. 26, 2024
Abstract
Efficient
algorithms
are
needed
to
segment
vasculature
in
new
three-dimensional
(3D)
medical
imaging
datasets
at
scale
for
a
wide
range
of
research
and
clinical
applications.
Manual
segmentation
vessels
images
is
time-consuming
expensive.
Computational
approaches
more
scalable
but
have
limitations
accuracy.
We
organized
global
machine
learning
competition,
engaging
1,401
participants,
help
develop
deep
methods
3D
blood
vessel
segmentation.
This
paper
presents
detailed
analysis
the
top-performing
solutions
using
manually
curated
Hierarchical
Phase-Contrast
Tomography
human
kidney,
focusing
on
accuracy
morphological
analysis,
thereby
establishing
benchmark
future
studies
within
phase-contrast
tomography
imaging.
Язык: Английский
A general strategy for generating expert-guided, simplified views of ontologies
bioRxiv (Cold Spring Harbor Laboratory),
Год журнала:
2024,
Номер
unknown
Опубликована: Дек. 18, 2024
Abstract
Annotation
with
widely
used,
well-structured
ontologies,
combined
the
use
of
ontology-aware
software
tools,
ensures
data
and
analyses
are
Findable,
Accessible,
Interoperable
Reusable
(FAIR).
Standardized
terms
synonyms
support
lexical
search.
Ontology
structure
supports
biologically
meaningful
grouping
annotations
(typically
by
location
type).
However,
there
significant
barriers
to
adoption
ontologies
researchers
resource
developers.
One
barrier
is
complexity.
Ontologies
serving
diverse
communities
often
more
complex
than
needed
for
individual
applications.
It
common
atlases
attempt
their
own
simplifications
manually
constructing
hierarchies
linked
but
these
typically
include
relationship
types
that
not
suitable
annotations.
Here,
we
present
a
suite
tools
validating
user
against
ontology
structure,
using
them
generate
graphical
reports
discussion
views
tailored
needs
HuBMAP
Human
Reference
Atlas,
Developmental
Cell
Atlas.
In
both
cases,
validation
source
corrections
content
hierarchies.
Язык: Английский
Atlases galore: where to next?
Nature Methods,
Год журнала:
2024,
Номер
21(12), С. 2203 - 2208
Опубликована: Дек. 1, 2024
Язык: Английский
Discovery of optimal cell type classification marker genes from single cell RNA sequencing data
bioRxiv (Cold Spring Harbor Laboratory),
Год журнала:
2024,
Номер
unknown
Опубликована: Апрель 27, 2024
Abstract
The
use
of
single
cell/nucleus
RNA
sequencing
(scRNA-seq)
technologies
that
quantitively
describe
cell
transcriptional
phenotypes
is
revolutionizing
our
understanding
biology,
leading
to
new
insights
in
type
identification,
disease
mechanisms,
and
drug
development.
tremendous
growth
scRNA-seq
data
has
posed
challenges
efficiently
characterizing
data-driven
types
identifying
quantifiable
marker
genes
for
classification.
machine
learning
explainable
artificial
intelligence
emerged
as
an
effective
approach
study
large-scale
data.
NS-Forest
a
random
forest
learning-based
algorithm
aims
provide
scalable
solution
identify
minimum
combinations
necessary
sufficient
capture
identity
with
maximum
classification
accuracy.
Here,
we
the
latest
version,
version
4.0
its
companion
Python
package
(
https://github.com/JCVenterInstitute/NSForest
),
several
enhancements
select
gene
exhibit
highly
selective
expression
patterns
among
closely
related
more
perform
selection
atlases
millions
cells.
By
modularizing
final
decision
tree
step,
v4.0
can
be
used
compare
performance
user-defined
computationally-derived
based
on
classifiers.
To
quantify
how
well
identified
markers
desired
pattern
being
exclusively
expressed
at
high
levels
within
their
target
types,
introduce
On-Target
Fraction
metric
ranges
from
0
1,
1
assigned
are
only
not
cells
any
other
types.
outperforms
previous
versions
ability
higher
values
approaches
significantly
F-beta
scores
when
applied
datasets
three
human
organs
-
brain,
kidney,
lung.
Язык: Английский
Discovery of optimal cell type classification marker genes from single cell RNA sequencing data
Deleted Journal,
Год журнала:
2024,
Номер
1(1)
Опубликована: Ноя. 4, 2024
Abstract
Background
The
use
of
single
cell/nucleus
RNA
sequencing
(scRNA-seq)
technologies
that
quantitively
describe
cell
transcriptional
phenotypes
is
revolutionizing
our
understanding
biology,
leading
to
new
insights
in
type
identification,
disease
mechanisms,
and
drug
development.
tremendous
growth
scRNA-seq
data
has
posed
challenges
efficiently
characterizing
data-driven
types
identifying
quantifiable
marker
genes
for
classification.
machine
learning
explainable
artificial
intelligence
emerged
as
an
effective
approach
study
large-scale
data.
Methods
NS-Forest
a
random
forest
learning-based
algorithm
aims
provide
scalable
solution
identify
minimum
combinations
necessary
sufficient
capture
identity
with
maximum
classification
accuracy.
Here,
we
the
latest
version,
version
4.0
its
companion
Python
package
(
https://github.com/JCVenterInstitute/NSForest
),
several
enhancements
select
gene
exhibit
highly
selective
expression
patterns
among
closely
related
more
perform
selection
atlases
millions
cells.
Results
By
modularizing
final
decision
tree
step,
v4.0
can
be
used
compare
performance
user-defined
computationally-derived
based
on
classifiers.
To
quantify
how
well
identified
markers
desired
pattern
being
exclusively
expressed
at
high
levels
within
their
target
types,
introduce
On-Target
Fraction
metric
ranges
from
0
1,
1
assigned
are
only
not
cells
any
other
types.
outperforms
previous
versions
simulation
studies
ability
higher
values
real
data,
approaches
significantly
F-beta
scores
when
applied
datasets
three
human
organs—brain,
kidney,
lung.
Discussion
Finally,
discuss
potential
cases
genes,
including
designing
spatial
transcriptomics
panels
semantic
representation
biomedical
ontologies,
broad
user
community.
Язык: Английский
Construction, Deployment, and Usage of the Human Reference Atlas Knowledge Graph for Linked Open Data
bioRxiv (Cold Spring Harbor Laboratory),
Год журнала:
2024,
Номер
unknown
Опубликована: Дек. 23, 2024
Abstract
The
Human
Reference
Atlas
(HRA)
for
the
healthy,
adult
body
is
developed
by
a
team
of
international,
interdisciplinary
experts
across
20+
consortia.
It
provides
standard
terminologies
and
data
structures
describing
specimens,
biological
structures,
spatial
positions
experimental
datasets
ontology-linked
reference
anatomical
(AS),
cell
types
(CT),
biomarkers
(B).
We
introduce
HRA
Knowledge
Graph
(KG)
as
central
resource
v2.2,
supporting
cross-scale,
queries
to
Resource
Description
Framework
graphs
using
SPARQL.
In
December
2024,
KG
covered
71
organs
with
5,800
AS,
2,268
CTs,
2,531
Bs;
it
had
10,064,033
nodes,
171,250,177
edges,
size
125.84
GB.
comprises
13
Digital
Objects
(DOs)
Common
Coordinate
Ontology
standardize
core
concepts
relationships
DOs.
(1)
provide
code
construction;
(2)
detail
deployment
Linked
Open
Data
principles;
(3)
illustrate
usage
via
application
programming
interfaces,
user
products.
A
companion
website
at
https://cns-iu.github.io/hra-kg-supporting-information
.
Язык: Английский