►
From YouTube: DevoWorm (2023, #5): DevoWorm on Huggingface, Modeling Autocatalysis, Morphogenetic Collectives
Description
DevoLearn/DevoWormML on Huggingface. Mathematical modeling of spontaneous autocatalysis in vesicles. Long-range interactions in morphogenesis and embryo networks, interactions within collectives of cells and extracellular matrix. Attendees: Richard Gordon, Susan Crawford-Young, Sushmanth Reddy Mereddy, Harikrishna Pillai, Gautham Krishnan, and Bradly Alicea
B
B
E
Yeah
yeah,
okay,
so
I'm
Gautam,
Krishnan
and
I'm,
a
sophomore
studying
for
a
B
Tech
CS
degree.
So
I
came
to
know
about
this
by
while
going
to
the
g-soft
projects
and
I
found
this
project
of
Devo.
Really
interesting
and
I've
had
some
experience
working
with
fight
torch,
computer
vision
models
and
what
is
very
interesting,
yeah.
D
E
D
A
Project,
so
that's
that'll
be
great.
You
know
we'll
keep
working
on
it.
I
know
we
didn't
put
together
a
gsoc
project
for
it
this
year,
but
you
know
it's
okay,
we
can
still
we
got
exceedingly
lucky.
Last
year
we
got
all
of
our
projects
funded,
I,
don't
know
what
it'll
be
like
this
year,
but
I
didn't
want
to
do
the
I
didn't
want
to
do
the
micro
series
project
again,
I
didn't
have
enough
bandwidth
for
it,
but
at
least
to
do
a
gsort
project.
We'll
still
do
some
things
with
updating
it.
A
Now
we
saw
we
have
the
two
digital
microsphere
platforms
that
we
need
to
kind
of
find
a
way
to
put
together
some.
You
know
Advance,
some
of
that
work
and
that's
usually
the
biggest
challenge
from
gsoc
this
year
we're
doing
a
gnn's
project
where
we're
doing
like
you
know
we're
we're
kind
of
updating
the
Pipeline
and
we're
integrating
it
with
Diva
learn.
A
It
just
seemed
like
the
lowest
hanging
fruit
from
the
standpoint
of
our
existing
Devo
learn
initiative.
So
I'm
still
interested
very
much
interested
in
doing
things
with
the
digital
microsphere
I,
don't
know
what
people
have
in
mind.
I
know:
Susan
probably
has
a
stake
in
this
as
well,
where
we
can,
you
know
Advance
some
of
the
capabilities
of
the
software.
So
right
now
it
just
takes
a
you
know,
captured
a
number
of
images
and
you're
tiling
the
sphere,
we're
doing
it
using
different
algorithms
and
we
can
do
machine
wooden
row.
A
D
C
What
are
the
models?
We
have
actually
I
wrote
the
dot
py
files
for
everything
for
every
model.
Actually,
I
want
to
combine
completely
with
like
every
model
in
one
page,
but
hugging
face.
Wasn't
so
good
thing
for
completely
flying
three
models
at
a
time
in
spaces,
so
I
need
to
deploy
after
separately.
I
have
actually
deployed
in
and
I
show
like
I
will
show
some.
What
I
get
I'll
just
I
will
do
and
then
I
will
deploy
into
yeah.
Here
are
the
models
actually
I
kept
them
private?
A
C
Now
we
are
getting
centroids
of
every
image
and
segmentation
map
of
it.
Actually,
these
are
for
images.
Only
I
was
thinking
to
implement
for
video,
also
like
a
Time
Lapse
video.
If
I
give
an
input
as
some
video
the
time
lapse
of
segmentation
map
and
centroid
map
will
be
generated,
and
these
can
be
saved
in
some
CSV
file.
I
was
thinking
to
work
and
deploy
in
the
actually
last
year
from
what
are
you
cover
coming?
C
I,
really,
don't
remember
in
his
name,
but
he
did
this
method
for
collecting
centroids
of
cells
for
videos
also
that
so
I
need
to
implement
that
I
was
thinking
to
deploy
them
so
actually
Brandy.
What
you
have
shared
to
me.
The
account
was
a
personal
account
like
me,
but
you
need
to
create
an
organization
like
from
here
fill
out
these
details.
If
you
create
an
organization,
then
I
can
deploy
them.
There,
I
mean
contribute
to
that
account.
Yeah.
C
E
C
C
Machine
learning,
so
I
was
thinking
to
implement
one
paper
actually.
Last
week,
I
have
given
a
presentation
right,
yeah
I
mean
before
last
week,
yeah
yeah
from
that
paper.
I
want
to
extract
every
area
of
each
cell,
and
so
what
we
call
curvature
of
itself
and
I
will
give
them
as
an
load
features
for
every
graph
neural
network.
There
will
be
more
features
and
every
node
will
be
connected.
Sorry
Edge
features
will
be
there
and
every
node
will
be
the
centroid.
C
Okay
in
parallel
and
connections
can
be
made
from
that.
Maybe
I
will
next
week
what
all
my
ideas
are.
There
I
will
just
give
a
small
glimpse
to
your
small
type
of
pizza
type
of
thing,
and
this
I
will
contribute
by
today
and
in
within
two
days,
I
mean
within
my
next
week.
I
will
try
to
implement
that
also
video
of
the
web
deployment
of
video.
If
we
give
a
video
see
you
again,
video
yeah,
it
will
generate
some
products
and
I
will
I
was
hoping
to
create
pipeline
for
this
web
development.
C
Also
Pump
It
Up
directly.
Every
time
someone
contributes
it,
it
loads
the
whole
thing
and
shows
on
in
plugging
case
actually,
and
it
cuts
it
barely.
These
are
things.
I
immediately
took
my
Taurus
to
write
this
whole
thing.
Actually,
I
was
so
I
I
have
seen
that's,
but
it's
not
working
so
I
started
writing
from
the
scratch.
First
I
made
lineage
population
now
after
that,
other
two
also
done
by
two
down
below
oh
yeah.
That's
great
yeah,.
A
C
C
C
B
A
C
That's
one
thing,
and
this
example
image
I
have
given
two
images,
and
this
is
another
image
see
segmentation
map
is
generated
here
we
can
save
them
actually,
I
have
provided
this
feature
Searcher
for
this
image.
We
can
download
the
image
download,
PNG
image.
I
can
zoom
it
or
after
scanner,
and
this
is
done
for
nuclear
segmentation
and
for
the
lineage
population.
C
F
C
E
C
F
A
C
A
C
F
C
A
This
is
a
little
bit
better
than
we
have
with
so
for
background
on
this
we
were
hosting
these
models
on
on
Heroku
app,
which
is
a
hosting
service.
Yeah
they've
recently
changed
their
conditions
where
you
know
you
have
to
pay
for
access
to
host
things,
and
so
that's
not
great,
and
so,
but
it
was
also
less
flexible
than
this.
So
hugging
faces
is
like
built
for
sharing
models,
I
think
and.
D
E
C
C
C
E
A
We
also
have
some
other
models
in
I.
Don't
I
know
you
probably
haven't
seen
this,
but
there's
Diva,
worm,
Ai
and
I.
Think
they're
on
GitHub
in
Divo
learn
repository
it's
in
a
different
repository.
It's
a
new
organization
and
those
are
you
know
those
are
something
we
can
put
up
there
as
well.
Do
you
need?
What
do
you
need
for
me?
You
need
me
to
have
like
a
organizational
account
made
ready
to
go
or.
C
C
A
We
also
need
to
make
you
a
collaborator
on
the
GitHub
organization,
yeah,
yeah,
okay,
and
then
that
looks
great
yeah.
Well,
I
will
keep
going
through
this.
Hopefully
this
looks
great.
This
is
kind
of
nice
to
have
like
someone
to
take
it
up,
and
you
know-
and
we
have
like
a
number
of
people
contributed
to
that
organization
over
time.
So
you'll
join
that
group
of
people
and
yeah.
C
D
C
A
A
A
D
A
Women
doing
some
maintaining
and
thank
you
for
that,
but
you
know
it's
it's
something
that
you
have
to
both
have
a
really
good
hand
in
the
organization
and
like
it's
just
kind
of
a
volunteer
thing
where
you're
trying
to
do
it.
You
know
it,
but
it's
like
it's,
not
your
full-time
job.
So
it's
always
a
challenge
to
keep
up
to
date.
On
that.
C
A
B
B
If
you
support
it
with
the
rest
of
the
tissue,
it
stands
up
and
if
you
support
it
on
all
sides,
it
looks
like
a
nice
hexagon.
If
you
support
it
by
both
sides,
it
becomes
long
it's
nice
playing
with
a
model
because
you
can
see
Its
Behavior
better
than
guests
by
gosh
on
the
computer
like
I,
have
this
model
on
the
computer,
but
it
keeps
crashing
so
I
assume
that's
crashed,
but
I
don't
think
so.
This
is
just
the
flat
version
of
it.
So.
B
B
F
B
B
B
E
B
F
A
B
Basically,
not
with
a
trust
that
was
not
not
on
this,
particularly
it's
console
I'm
using
and
I'm
using
a
truss
model,
and
that
does
not
change
with
scale
which
I
think
is
wrong,
but
they
haven't,
including
the
nanotech
in
the
truss
model,
by
going,
why
not
and
they,
but
nobody
builds
a
building.
That's
a
nano
size.
D
B
B
A
A
F
B
B
B
B
A
A
B
F
B
D
F
B
B
F
A
B
E
A
A
A
B
B
F
F
Okay,
okay,
but
I'm,
saying
the
the
microtubules
have
to
grow
to
valence
of
the
cell,
but
they're
in
Dynamic,
instability
and
they're
forming
a
column
display
a
bundle
of
microtubules
because
they
might
be
crosswind
okay,
but
they
might
individually
be
changing
in
length.
So
how
stable
is
how
stable
is
just
the
structure
which
holds
the
cell
elongated.
B
A
A
Something
I've
been
working
on
with
dick
and
I'm
still
not
there
yet,
but
I
want
to
go
over
some
of
the
code.
So
I
do
this
in
scilab,
which
is
it's
like
Matlab,
but
it's
an
open
source
version
of
Matlab.
So
it's
got
like
a
lot
of
the
same
commands
as
Matlab
and
you're.
Manipulating
matrices
to
you
know,
do
modeling
and
things
like
that,
but
so
the
the
language
is
somewhat
opaque.
A
B
C
A
Yeah,
so
it's
kind
of
hard
to
get
so
this
is
the
I
mean
this
is
very
hard
to
follow,
because
it's
there
are
a
lot
of
edits
in
here.
But
basically
the
problem
is
is
that
we
have
this
vesicle
and
we
have
all
these
molecules
out
in
the
world
and
they're
coming
into
the
vesicle
and
they're.
The
the
membrane,
which
is
the
edge
of
the
vesicle,
has
a
certain
permeability,
and
so
the
molecules
will
come
into
the
vesicle
and
go
out
of
the
vesicle
and
at
the
beginning
of
Life.
A
Apparently,
the
vesicle
is
very
permeable,
meaning
that
you
could
take
in
a
large
number
of
molecules
and
keep
them
in
this
vesicle,
and
then
they
would
also
leak
out
at
a
very
high
rate.
So
you
know
the
idea
is
that
they
can't.
You
know
you
get
like
a
certain
number
of
types
of
say.
You
have
a
lot
of
molecules
and
of
the
that
large
collection
of
molecules.
A
You
have
a
certain
number
of
types,
and
so
you
have
to
sample
from
a
large
pool
molecules
a
certain
number
of
types
and
if
you
sample
enough
of
them,
you
can
get
them
into
the
mount
into
the
vesicle
and
you
get
like.
Maybe
all
the
types
you
need
to
start
Auto
catalysis
within
the
vesicle
and
the
their
conditions
on
this
like
at
night,
the
membrane
gets
blocked
by
large
macromolecules,
and
so
things
can't
leave
the
the
vesicle.
But.
A
A
And
so
the
the
relation
to
like
things
like
neurons
and
modern
cells
is
that
modern
cells
have
these
ion
channels,
So.
Eventually
the
edges
of
the
the
vesicle
would
be.
You
know
the
edge
of
the
vesicle
or
the
lipid
membrane
would
become
a
formal
cell
membrane,
a
bilayer
membrane
and
then
they'd
have
these
ion
channels,
which
would
let,
in
certain
molecules
at
a
in
a
controlled
way
that
you
know
result
in
in
Electro
electrical
potential
and
things
like
and
other
types
of
things
that
the
cell
needs.
A
F
A
A
So
the
the
membrane
channels
are
the
only
way
things
can
get
in
and
out.
The
rest
of
that
membrane
is
is
not
letting
anything
through.
So
it's
a
very
selective
sampling
of
the
environment,
so
you
know
for
ION
channels.
It
only
lets
in
one
type
of
ion
or
one
class
of
ion
like
a
potassium
ion
or
something,
and
it's
very
much
specialized
for
that.
So
you
don't
get
a
bunch
of
things
in
there.
A
A
So
I
said
before
that
everything
can
come
in
and
everything
can
go
out.
That's
not
entirely
true!
You
can
have
you
know,
maybe
a
little
bit
of
infermeability.
Of
course,
at
night
you
get
zero
in
per
zero
zero
Perma
permeability
during
the
daytime.
It's
one
permeability
value
of
1.0
because
there's
no
impermeability,
so
we
can
average
that
over
a
daytime
like
an
entire
day
length
as
0.5.
A
So
that
would
be
like
a
baseline
case
and
then
it's
it's
it's
homogeneous
for
in
and
out
flux,
so
influx
and
out
flux,
don't
vary
so
like
in
an
ion.
Channel
influx
and
outflix
very
much
would
vary
or
you
could
have
a
situation
where,
if
you
wanted
to
see
what
happens,
you
know
it
it
day
versus
night.
You
could
change
those
values,
but
this
is
again
the
standard.
A
Then
you
know
we're
just
drawing
from
a
distribution
here.
So
this
line
here
is
where
you're
drawing
you
know:
you're
you're,
creating
a
sample
in
the
environment
of
all
these
different
molecular
kinds
and
molecules
which
are
basically
numbers
and
it's
just
generating
a
series
of
numbers.
It's
generating.
You
know
a
distribution
of
different
numbers,
and
then
you
know
you
have
the
kinds
which
are
like
a
single
number
and
then
there
are
different
copies
of
that
in
the
environment
and
then
you're
sampling
from
that
distribution.
A
And
then
you,
you
know
you
have
a
histogram
of
whatever
you're
capturing.
So
the
more
you
sample
you
know,
the
larger.
The
C
value
is
the
more
you
sample
and
the
greater
you
can,
the
greater
number
of
kinds
you
can
capture
at
any
one
time.
So
this
is
just
you
can
generate
a
histogram
of
of
all
the
different
types.
F
F
D
F
Okay,
the
basic
idea
is
when
you
get
a
certain
group
of
molecules
together.
That,
though,
are
they
become
a
metabolism
and
the
cell
becomes
alive
okay.
So
it's
it's
sort
of
a
a
simple.
My
simple
model
for
what
makes
something
alive
and
all
you
have
to
do-
is
get
get
a
sample
in
this
case
36.
If
you've
got
at
least
one
molecule
of
36
different
kinds
inside
this
inside
the
vesicle,
then
it
becomes
alive.
A
A
A
A
A
A
Had
a
version
of
this
so
there's
eigen
and
often
and
that
basically
they
propose
that
you
can
have
this
set
of
molecules
that
are
sort
of
an
association
with
one
another
and
they
form
the
causes
hyper
cycle.
So
it's
like.
If
you
have
all
these
molecules
in
place,
they
can
form
a
reaction
or
a
chain
reaction,
and
it
can
form
this
cycle,
then
that
reinforces
itself.
A
Yeah
yeah
yeah.
So
that's
so
that's
that's
where
that
is
and
then,
if
we
go
through
so
then
we
can
do,
we
can
do
some
checks
with
histograms
and
things
to
see
where
we
are
and
then
we
end
up
with
this.
Well,
we
can
do
a
subtractive
model
if
we
take
the
outflux,
but
that
if,
if
both
of
the
values
are
influx
and
out
flux,
they're
the
same-
that's
not
really
important,
but
it
yields
us
this
sort
of
Stoichiometry
of
different
molecules
in
the
in
the
vesicle.
A
So
a
Stoichiometry
is
like
a
sort
of
a
level
of
certain
molecule
in
a
Cell
they'll
use
it
in
molecular
analysis.
If
it's
like
a
very
small
molecule,
molecular
or
number
of
molecules
versus
a
very
large
number
of
molecules,
and
they
use
that
in
like
stochasticity
analysis
where
they
look
at,
like
you
know,
different
type,
specific
types
of
molecules,
if
they're
a
very
small
number
of
very
large
number-
and
they
can
do
different
types
of
things
of
that.
A
But
that's
that's
where
that's
what
the
code
does
and
then
so
they
generate
something
like
this,
so
this
is
for
two
days.
This
is
showing
the
number
of
Kinds.
This
is
for
this
value
or
actually
these
three
lines.
The
blue
line
is
for
vesicle
diameter
of
40..
The
green
line
is
for
a
vesicle
diameter
of
100
and
the
red
line
is
a
vesicle
diameter
of
200,
so
it
increases
and
remember
our
36
kind
threshold,
which
is
here.
A
This
blue
line
is
the
day
night
cycle,
so
we
assume
that
the
day
cycle
has
you
know:
you'll
have
all
36
kinds
in
the
cell
and
then
they
leave
the
or
they
get
trapped
in
the
Cell
at
night
or
actually,
okay.
So
during
the
day
the
number
of
Kinds
is
36
and
then
at
night
it
switches
off
so
that
you
don't
get
an
influx.
E
A
F
C
E
F
A
A
So
if
you
have,
if
you
have
a
lucky
end
of
the
day
where
you
get
a
lot
of
different
kinds
and
you
get
like
elevated,
but
then
you
lose
that-
and
you
can
see
that
like
so
you
can
see
that,
like
for
the
200
size,
you
start
to
get
to
that
threshold,
and
you
know
this
is
this
is
for
two
days.
So
this
is,
you
know,
could
be
simulated
over
in
this.
A
F
A
Yeah
yeah,
so
this
is
fluctuating,
and
so
this
is
actually
where
I
read
so
I
ran
this
idea,
where
you
have
this
mix
of
a
uniform
and
a
poisson
distribution,
and
then
this
is
the
result,
it's
kind
of
hard
to
see,
but
this
is
for
10
to
the
fifth
days.
So
this
is
going
like
on
the
daily
average
10
to
the
fifth
days.
I
think
it's
yeah!
A
You
get
this
you
you
go
over
the
36
kind
threshold
quite
regularly,
and
so
it
tells
you
that
you
know
you
can
exceed
that
threshold.
If
you
have
this
sort
of
structured
environment
a
little
bit,
it
doesn't
have
to
be
very
structured.
If
you
do
like
a
exponential
distribution,
you
consistently
go
over
the
36
threshold.
So
it's
you
know
it's
the
structure
and
the
exponential
distribution
would
just
simply
be
a
more
non
non-normal
or
non-random
case
where
you
have
a
more
localized
structure.
F
F
A
Right
well
so
the
rate
is
like
the
diameter
is
basically
the
size
of
the
cell,
so
the
idea
would
be
that
you
can
get.
Is
you
can
get
stuff
into
the
cell?
And
then
you
know
at
some
point:
it's
just
going
to
be
like
crowded
enough
where
you
can't
get
any
more
in
so
the
diameter
of
the
cell
is
like
the
maximum
sampling.
So
basically.
A
Yeah,
but
you
have
like,
if
you
have
a
diameter
of
a
hundred,
you
can
sample
100
things,
and
so
what
are
the
chances
of
getting
36
kinds
of
a
random
sample
of
a
hundred
and
that's
somewhat
low?
But
if
it's
bigger,
then
you
get
more
molecules
in
there.
You
can
get
two
or
three
hundred
and
then
that's
right
closer
to
36,
because
when
you
sample
a
number
of
Kinds
you
get
a
large.
A
A
A
A
F
It
possibly
has
to
do
with
proteins,
embedded
or
peptides,
embedded
in
the
cell
membrane,
which
don't
fit
precisely,
and
also
possibly
remnants
of
the
leakiness
of
the
initial
numbers,
because
they're
making
they're
made
of
a
whole
mix
of
ampophiles,
so
they
probably
aren't
leaky,
and
so,
as
the
leakiness
decreases
for
the
membrane,
the
peptides
in
the
membrane
may
bring
it
back
up.
That
may
make
it
selective
okay,
but
that's
as
far
as
I
can
go
right.
Now,
if
you'd
like
to
learn
molecular
Dynamics,
then
we
can
do
some
simulations
yeah.
D
D
D
E
F
A
D
C
D
A
Okay,
so
yeah
our
gstock
projects,
like
Garner,
posted
the
incfs
to
go
through
a
process
of
being
approved
by
the
gsoc
program,
but
they
usually
do
so,
but
the
projects
are
listed
instead
of
the
gnm's
project
is
formally
listed
up
on
their
website,
so
we're
starting
to
get
people
to
come
in
on
this,
and
so
I
mean
you
know
the
the
application
processes.
You
know
it's
it's
it's
a
time
where
people
like
to
contribute
to
the
Oregon
right.
C
Do
you
have
any
idea
like
actually
for
writing
proposal
I
completely
shot
down?
My
college
and
I
came
to
home
to
write
it
for
next
20
days,
but
so
I
was
thinking
to
work
on
some
small
things.
If
you
have
any
ideas
or
any
papers
could
before
the
before
before
project,
if
I
want
to
have
any
knowledge,
can
you
share
it
not
graph
your
own
network,
because
what
all
we
have
is
microscopic
videos
right,
microscopic
videos
and
relationship
to
GM,
and
if
you
have
any
papers
or
resources,
please
share
it
to
me.
A
A
C
C
A
Thank
you,
bye,
see
you
later
bye,
see
you
later
bye.
All
right
now,
I'd
like
to
talk
about
a
few
papers
in
our
reading
queue
that
I'd
like
to
clear
out
it's
an
assortment
of
topics.
So
the
first
one
is
reciprocal:
cell
extracellular
Matrix
Dynamics
generates
supracellular
fluidity
underlying
spontaneous
follicle
patterning.
So
we
talked
about
this
during
the
push
pull
morphogenesis
paper
series
that
we
had
last
year
in
2022,
and
then
Susan
talked
about
this
in
the
context
of
tensegrity
networks,
I
think
last
week
in
her
or
the
week
before
in
her
presentation.
A
A
They
can
show
that
you
can
do
this
on
a
plating
in
a
plated
context,
so
you
can
take
the
cells
from
dermal
tissue,
dissociate
them
and
then
plate
them
in
in
a
number
of
these
small
Wells.
So
this
is
a
96
well
plate
and
then
for
each
well,
you'll
see
the
cells
and
a
form
these
Rings,
instead
of
just
flat
Aggregates
or
these
dermal
bumps,
and
so
you
can
see
the
rings
that
come
that
form
in
place
of
that.
A
And
so
you
see
that
there's
an
aligned,
a
gradual
alignment
with
the
pattern
axis
which
is
kind
of
this
ring
area
here,
and
then
you
get
an
ordered
cell
ECM
layer
from
that,
and
then
you
get
these
periodic
multicellular
Aggregates,
which
are
actually,
if
you
go
further
out
from
the
formation
of
that
ring
and
they
actually
look
like
now.
They
actually
look
like
these
multicellular
Aggregates,
these
follicles
on
the
surface
of
this
embryonic
skin
layer.
A
So
you
can
basically
replicate
this
in
culture.
You
can
show
you
know,
pattern
formation
from
Aggregates
to
a
ring
to
these
clusters
of
follicles,
and
so
this
is
something
that
we
can
look
at
this
whole
process
over
48
hours
in
the
xvivo
example,
and
then
you
can
see
that
there
are
these
contractile
cells
that
align
the
extra
cellular
matrix.
This
is
in
Vivo.
The
aligned
extracellular
Matrix
increases
contractility
through
this
ion,
calcium,
ions
and
then
contractile
and
stability
is
where
you
get
this
formation
of
these
clusters
they're
an
instability
of
contractile
forces.
A
A
So
this
is
the
nice
people,
so
this
paper
kind
of
goes
over
this
model
in
avian
Skin
So
during
embryogenesis
cell
collectives,
engage
in
coordinated
Behavior
to
form
tissue
structures
of
increasing
complexity
in
the
avian
skin
assembly
into
follicles
depends
on
intrinsic
mechanical
forces
of
the
dermis,
but
how
cell
and
mechanics
initiate
pattern
formation
is
not
known
here
we
constitute
the
initiation
of
follicle
patterning
X
Vivo,
using
only
freshly
dissociated
Avion,
dermal
cells
and
collagen.
So.
A
This
this
chicken
model
they're
using
the
cells
from
the
skin
of
the
chicken
they're,
also
looking
at
the
skin
formation
in
the
embryo,
we
find
that
contract
our
cells
physically
rearrange
the
extracellular
Matrix
and
the
ECM
rearrangements
further
align
cells.
So
this
is
this
extracellular
Matrix.
This
is
outside
of
the
cells.
A
You
have
extracellular
Matrix
at
different
densities,
so
we
saw
in
Susan's
presentation
where
she
was
modeling,
not
just
the
cells
but
the
extracellular
Matrix
in
the
orientation
of
of
different
things
in
that
Matrix
in
relation
to
the
cell
membrane,
in
some
of
the
actin
molecules
in
the
cell
membrane.
So
this
is
an
interact,
a
series
of
interactions
and
they
transmit
mechanical
forces
not
only
within
cells
but
between
cells
as
well.
A
This
exchange
transforms
a
mechanically
unlinked
Collective
of
dermal
cells
into
a
Continuum
with
coherent
long-range
order.
So
this
is
something
that
you
see
a
lot
in
complexity,
Theory,
coherent
long-range
order.
It's
basically
something
you'll
get
out
of
non-linear
interactions.
So
when
they're
non-linear
interactions,
you
tend
to
get
these
long-range
order
effects.
Things
are
connected
together,
they're
kind
of
influencing
one
another.
At
the
you
know,
the
nearest
neighbor
level,
and
then
you
get
these
long
range
effects
that
are
coordinating,
and
so
we
can
talk
more
about
that
in
theory
or
later.
A
But
for
now
we
showed
that
this
ordered
cell
ECM
layer
behaves
as
an
active
contractile
fluid
that
spontaneously
forms
regular
patterns.
So
this
whole
thing
is
like
a
fluid.
Instead
of
you
know
you
think
of
a
tissue,
maybe
as
more
of
a
solid,
but
it
actually
behaves
more
like
a
fluid,
and
so
it
spontaneously
forms
circular
patterns,
which
is
what
we
see
in
in
orthogenesis
in
a
lot
of
these
patterns.
A
Our
study
illustrates
a
role
for
mesenchymal,
Dynamics
and
generating
cell
level
ordering
and
tissue
level
patterning
through
a
fluid
instability
process
that
may
be
at
play
across
more
morphological
symmetry
breaking
contexts.
So
if
we
compare
this
with
the
reaction
diffusion
model
of
the
Turing
model
of
morphogenesis,
we
see
some
parallels
with
some
of
the
dynamic
instabilities
that
you're
looking
at
and
the
role
of
morphogens.
A
So
if
you
go
down
further
in
the
paper,
they
Identify
some
molecular
factors
that
are
at
play
here,
but
they
also
identify
the
geometry
and
some
of
the
aspects
of
the
geometry
with
respect
to
the
cells
and
the
extracellular
Matrix.
So
this
is
maybe
an
application
domain
for
integrity
networks
we're
pulling
in
some
really
interesting
applications
of
Turing
morphogenesis.
A
So
this
is
where
we
show
the
dermal
cells
display
supercellular
order
preceding
feather
follicle
emergence,
that's
a
mouthful,
but
basically
this
is
the
alignment.
And
then
this
is
the
pattern.
So
you
see
in
in
panel
a
you
see
this
alignment
of
cells,
this
initial
alignment,
and
then
you
see
the
patterning
in
F,
where
you
start
to
get
these
clumps
that
look
like
these,
this
discrete
pieces
of
the
followable.
So
you
have
this
these
stains
for
actin.
A
You
have
the
nucleus
stains
and
you
have
collagen
and
fibronectin
stains
and
shows
where
those
are
in
the
culture.
And
then
you
can
see
the
actin,
the
player
role
so
and
B
and
C.
You
have
maximum
intensity
projections
of
effactin
and
the
nuclei
stained
in
dappy,
which
is
this
kind
of
stain
these
to
stay
nuclei.
It's
usually
a
blue
type
stain
and
the
epidermis
and
dermis.
So
this
is
these
two
panels
are
for
E6
and
e65.
These
are
just
developmental
stages
and
then
down
here.
A
This
is
for
actin
and
nuclei
in
in
F,
which
is
this
is
the
embryo
back
skin
at
six
point
or
it's
day,
seven,
and
then
this
is
day
6.5
in
a
so.
This
part
here
is
a
little
bit
earlier
than
this
part,
and
you
can
see
that
you
start
to
get
this
formation
of
like
this
alignment
at
day,
six
and
a65,
and
then
this
clustering
at
day.
Seven!
A
So
that's
and
then,
of
course,
this
is
depending
on
the
orientation
along
the
AP
access.
So
this
is
going
along
the
AP
access
and
E,
and
then
this
is
the
percent.
So
it's
like
the
percent
orientation
and
you
get
this
Arc
at
the
edge
here,
which
is
going
up
and
down
the
AP,
Axis
or
percent
of
orientation.
You
find,
and
you
find
that
in
one
part
of
the
AP
access,
I
guess
this
is
the
anterior
end.
Here
you
get
more
order
than
at
the
posterior
end.
A
So
it's
this
long-range
order
which
organizes
us
along
the
AP
axis.
It's
not
just
pairwise
interactions.
It's
this
tissue
level,
interaction
that
goes
all
the
way
up
and
down
the
anatomical
axis,
and
then
this
is
of
course,
for
day
six
and
day,
six
five.
So
you
can
see
that
in
day,
6.5
or
E
6.5
it's
a
lot.
It
gets
coordinated
a
lot
more
along
the
axis
than
in
D6
and
day
six.
It
kind
of
goes
across
to
about
half
the
axis,
and
then
it
goes
to
zero
percent
orientation.
A
I
guess
the
orientation
is
the
orientation
in
common.
But
the
point
being
here
is
that
there's
more
long-range
order
as
we
go
out
from
day
six
to
day
seven,
and
then
this
is
showing
the
alignment
and
pattern
process.
So,
at
day
six
you
have
this
alignment
at
day,
six
five,
you
have
alignment,
that's
refined
and
then,
by
day,
seven
you
get
these
clusters,
and
so
this
shows
that
the
cellular
level,
what
that
looks
like
and
they're,
showing
these
these
extracellular
Matrix
networks,
which
actually
map
very
nicely
onto
this
tensegrity
network
problem.
A
A
You
see
the
nucleus
stain
of
ACTA
in
the
nuclei
and
the
actin
stain
and
these
cells,
so
you
can
see
they're
organizing
into
these
rings,
and
we've
seen
this
before
in
different
papers,
where
you
get
these
Rings
of
acting
people
that
we
had
a
paper
where
they
did
this
in
a
vesicle.
So
this
is
something
that
happens
a
lot
not
just
in
in
chicken
morphogenesis,
but
this
shows
again
some
of
these
graphs.
So
this
is
the
percent
of
the
actin
molecules
in
parallel.
This
shows
time.
A
So
this
is
a
cellular
orientation,
so
you
can
see,
there's
a
peak
where
they're
coordinated
and
then
it
falls
off
after
40
hours,
which
is
this.
This
measure
that
they're
using
here
and
then
actin
orientation,
which
is
again
mimicking
this
thing
we
saw
in
the
embryo.
So
this
is
over
eight
hours,
20
hours
and
30
hours,
so
you
can
see
as
we
go
on
in
time.
We
get
this
circumferential
axis,
which
is
actually.
This
is
in
the
plated.
A
Well,
this
is
in
culture
versus
in
the
embryo,
so
you
don't
have
an
AP
axis
here.
You
just
have
the
circumference
of
the
dish
of
the
plate.
Well
and
then
the
cells
are
aligning
themselves
within
this,
so
you
shouldn't
expect
to
see
like
a
full
embryogenetic
transformation.
You
should
expect
to
see
these
rings
and
then
perhaps
these
individual
clusters
of
cells,
so
you
can
see
that
the
orientation
is
peaking
at
40
hours
and
then
it
falls
off
likewise.
A
A
In
particular,
we
find
that
our
ex-vivo
assay
cells
are
in
mysternycm,
which
is
the
extracellular
Matrix,
and
so
the
cell
ECM
layer
undergoes
a
reversible
Arrangements
on
the
time
scale
one
to
ten
hours.
So
this
is
in
culture,
where
they
have
this
time
scale
of
10
hours
where
these
rearrangements
go
on
and
it's
an
irreversible
Arrangement.
So
it's
not
something
that
just
falls
apart
with
you
know
if
you
treat
it
with
something
it
organizes,
and
then
it
falls
apart
in
the
absence
of
the
signal.
A
This
also
suggests
that
embryonic
chicken
skin
possesses
viscous
fluid-like
characteristics,
namely
it
continuously
deforms
or
flows
over
time
under
constant
applied
force,
and
it
retains
that
deformation
after
force
is
released.
So
this
is
not
you
know.
Skin
is
very,
has
this
sort
of
viscosity
as
a
sort
of
give-
and
you
know,
there's
a
certain
physics
to
skin,
but
in
the
embryonic
skin
it
actually
behaves
more
like
a
viscous
fluid
than
like
a
solid,
and
so
this
is
where
they're
getting
into
some
of
these
aspects
of
the
physics
of
it.
A
Studies
of
viscous
properties
and
tissues
are
underexplored
and
viscosity
is
often
excluded
in
favor
of
studies
of
elastic
Behavior.
So
this
adults,
their
mature
skin,
is
very
elastic
and
you
can
you
know
pinch
yourself
and
pull
it
out
and
it
has
to
be
elastic
because
it's
sitting
over
a
bunch
of
Limbs
and
muscles
that
move
around
a
lot,
so
it
has
to
have
a
lot
of
give.
But
in
the
embryo
it's
actually
more
like
a
viscous
fluid.
A
So
looking
at
viscosities
are
important
and
I
might
add
that
intensegrity
networks,
this
sort
of
macro-
you
know,
soft
matter.
Physics
is
probably
an
important
aspect
of
the
model,
because
you
know
we
have
these
models,
these
physical
models
of
tensority
that
involve
little
blocks
and
things
like
that.
A
We
expect
that
dermal
cells
act
as
attractive
elements
when
embedded
in
an
ECM
layer
drawing
in
adjacent
cells.
Three,
we
assume
that
the
number
of
in
spacing
of
Aggregates
is
determined
at
the
early
stages
of
the
patterning
process,
I.E
the
first
10
to
20
hours
in
Vivo.
The
density
patterns
are
read
out
at
e
7.5,
which
is
12
to
24
hours
after
a
patterning
onset
by
beta-katin
and
signaling.
A
It's
much
like
when
you
get
pattern
cell
death
or
a
patterned
apoptosis
when
you
have
like
limbs
on
a
or
fingers
on
a
limb
on
a
hand
limb.
So
you
have
these
cells
that
die
off
in
between
the
fingers.
You
start
out
with
a
webbed
hand,
and
the
cells
die
often
form
independent
fingers,
and
so,
but
this
is
also
with
more
of
a
molecular
turnover
model.
A
So
this
is
so.
Thus
we
aim
to
describe
patterning
onset
in
early
aggregation
and
neglect
efforts
such
as
spatial
variations
in
rheological,
Properties
or
dynamic,
parameter
changes
that
may
eventually
rise
from
feedback
between
cells
in
the
ECM
upon
strong
aggregation.
So
there's
a
lot
of
work
to
do
in
terms
of
modeling
of
this
type
of
thing.
A
A
This
paper
actually
fits
in
well
with
this
with
the
previous
paper
that
we
talked
about
the
we
talked
about
Push
Pull
morphogenesis,
and
this
paper
talks
about
pulsations
and
flows
and
tissues
as
two
Collective
Dynamics,
the
simple
cellular
rules.
So
in
the
last
paper
we
talked
about
long-range
order.
We
talked
about
it,
behaving
like
a
fluid
and
in
this
case
we're
talking
about
something
like
a
sort
of
fluid
metaphor,
so
we're
talking
about
pulsations
flows
of
pulsations,
our
pulses
in
time
there,
where
they
could
be
pulses
in
space
they're.
A
Just
quick,
bursts
of
things
and
flows
are,
of
course,
things
that
are
continuous.
They
have
a
rate
and
they
usually
move
across
space,
and
but
they
also
move
across
time.
So
these
are
dynamics
on
these
simple
cellular
matrices
that
we
see
here.
So
we
can
look
at
this.
This
is
the
extended
abstract
or
the
graphical
abstract.
They
have
the
pulsations,
which
are
marked
with
these
colored
markers.
A
There's
a
alignments
alignment
differentiation,
so
there
some
of
them
are
aligned.
Some
of
them
are
doing
other
things.
Then
you
see
here
where
they're
kind
of
curving,
so
you
see
that
the
layer
is
not
uniform,
but
it's
not
like
deformed
in
the
in
the
at
the
left
and
then
as
we
go
towards
the
right.
There
are
these
different
deformations
that
occur
as
a
result
of
these
pulses.
So
these
pulses
kind
of
organize
the
cell
layer
different
ways.
A
That's
just
general
pulsations,
now,
Focus
pulsations
are
these
things
where
they're
kind
of
moving
in
certain
directions,
and
you
can
see
that
they're.
So
these
two
pulses-
or
these
four
pulses
are
coming
together
to
construct
these
cells
in
the
middle
and
they
can
also
pull
cells
apart.
They
can
expand
the
area
between
them,
so
the
pulses
can
work
together
and
then
you
have
these
flows
which
are
over
time
specifically,
and
so
these
are
flowing
through
this
Matrix
and
it's
providing
some
organization
to
it.
So
it's
aligning
the
cells
according
to
the
flow.
A
This
is
actually
a
reminiscent
of
a
drosophila
Furrow,
where
you
have
this
Furrow
that
moves
across
the
omitidia,
which
has
a
imaginal
disc
in
drosophila
development,
and
it
aligns
the
cells
as
it
moves
across.
So
we
see
this
in
actual
morphogenesis,
but
this
just
shows
like
some
examples
of
this.
So
the
highlights
of
this
are
two
Collective
cell
motions.
Pulsations
and
flows
coexist
in
the
MD
CK
monolayers.
So
that's
what
they're
looking
at
here
ocean
is
controlled
by
the
regulation
of
substrate,
friction
and
cytoskeleton.
A
So
the
site
of
these,
these
physical
aspects
of
cells
are
actually
controlling
the
motion.
They're
feeding
back
to
the
Motions
you
get
motions
there
may
be
driven
by
external
forces,
but
they're
mediated
by
some
of
these
things
in
the
in
the
cell
interactions,
and
then
a
Vertex
model
recapitulates
the
motion
by
tuning
velocity
and
polarity
alignment.
So
you
get
this
vertex
a
model
which
I
think
I
think
is.
This
usually
has
like
vertices
in
the
network
as
well.
A
It
recapitulates
the
motion
by
tuning
in
polarity
alignment,
so
you
can
actually
recapitulate
some
of
these
movements
in
another
type
of
model.
So
summary,
here,
States
Collective
motions
of
epithelial
cells
are
essential
for
morphogenesis
tissues,
elongate
contract
flow
and
oscillate
the
sculpting
embryos.
So
you
get
from
this
ball
of
cells
to
this
highly
asymmetric
phenotype.
Through
these
kinds
of
processes,
these
tissue
level
Dynamics,
are
known,
but
the
physical
mechanisms
of
the
cellular
level
are
unclear.
A
Here
we
demonstrate
that
a
single
epithelial
monolayer
of
mdck
cells
can
exhibit
two
types
of
local
tissue,
kinematics,
pulsations
and
long-range
coherent
flows.
So
kinematics
are
emotions
they're,
you
know
distinct
from
Kinetics.
You
can
observe
them
with
tracking
dyes
and
microscope.
You
know
different
types
of
movements
of
things
that
are
stained
in
a
static
way
that
move
over
time.
So
you
can
measure
these
in
different
ways,
but
you
can
see
these
pulsations
and
long-range
coherent
flows
and
in
this
case
they're
characterizing
using
quantitative
live
Imaging.
A
We
report
that
these
motions
can
be
controlled
with
internal
One,
external
cues,
such
as
specific
Inhibitors
and
substrate
friction
modulation.
We
demonstrate
the
associated
mechanisms
of
the
unified
vertex
model
when
cell
velocity
alignment,
random
diffusion
of
cell
polarization
are
comparable.
A
positive
flow
emerges,
whereas
tissue
undergoes
a
long
range
flow.
When
velocity
alignment
dominates,
which
is
consistent
with
cytoskeletal
Dynamics
measurements,
we
propose
that
environmental
friction
ectomyosin
distributions
and
cell
polarization
kinetics
are
important
in
regulating
dynamics
of
tissue
morphogenesis.
A
A
These
changes
in
shape
are
reiterated
in
evolution
with
certain
families
of
tissue
transformation
such
as
the
elongation
and
gastrulation.
Among
many
morphogenetic
events,
even
if
their
genetic
backgrounds
are
different,
the
basic
cellular
form
mechanisms
are
expected
to
be
shared
among
different
biological
systems.
A
Indeed,
simple
physical
rules
to
achieve
morphogenesis
are
known
one
tissues
or
composed
of
cells
that
change
their
shapes
over
time.
Two
cells
and
tissues
interact
with
their
neighbors
by
adhering
to
each
other
via
adherent
injunctions
and
to
their
surrounding
extracellular
Matrix
their
focal
contacts.
So
this
is
just
the
way
that
the
cells
are
connected
to
one
another.
So
we
talk
about,
like
you
know,
nodes
and
edges
and
the
edges
are
these
focal
contacts
is
adherent.
A
A
A
A
So
during
blood
vessel
formation,
seven
events
were
reported
to
contribute
to
angiogenesis
and
it
was
the
types
of
different
events:
membrane,
Fusion
cell
migration,
cell
elongation,
rearrangements
cell
splitting
cell
cell
fusion
and
cell
division.
So
those
are
just
events
that
occur
during
that
process.
Each
of
them
may
contribute
to
the
formation
of
the
vessel
and
the
associated
cellular
flows.
So
this
General.
This
is
a
general
mechanism
called
Branch
formation,
which
is
essential
to
shape
organs
and
goes
beyond
the
formation
of
vascular
systems.
A
So
you
can
see
Branch
formation
and
all
sorts
of
organs
that
are
forming
through
a
morphogenesis
across
different
things
like
lungs,
kidneys,
mammary
glands.
They
just
have
different
ways
of
doing
this,
but
they
all
basically
do
this
branching
at
the
Single
Cell
level,
and
this
leads
to
Global
cellular
reorganizations,
so
they're
kind
of
getting
at
some
of
these
broader
sort
of
ways
of
doing
this.
A
Okay,
so
these
are
so.
This
is.
This
shows
them
doing
this
on
an
epithelial
monolayer,
showing
these
pulsations
and
flows.
So
these
are
the
pulsations
here
delivered
at
different
points
of
time.
These
lines
are
just
kind
of
the
pulsations
that
are
observed.
You
see
this
winding
index,
which
is
a
physical
measure,
physics
measurement
of
Dynamics.
A
So
you
can
see
this,
how
these
pulsations
sort
of
the
forces
and
how
they
move
around
the
surface,
the
monolayer,
and
then
you
get
these
flows
that
are
marked
in
these
different
colored
tracks,
and
then
you
see
the
winding
index
for
those
as
well,
so
the
flows
moving.
You
can
actually
track
sort
of
the
intensity
of
the
flow
using
this
winding
index
across
the
model.
Here
they
have
this
Divergence
and
velocity
magnitude
measurement
this.
A
This
heat
map,
which
shows
how
pulsations
and
flows
contribute
to
some
of
these
differences
in
velocity
and
differences
in
Divergence
across
the
monolayer,
and
then
you
have
this.
These
graphs
down
here
would
show
Divergence
and
velocity
magnitude
over
time,
and
this
just
shows
like
the
for
that
index-
that
it
kind
of
settles
down
as
time
goes
on.
So
there's
a
lot
of
activity
early
on
and
it
settles
down
over
a
48
hour
period,
foreign.
A
A
A
And
so
there
are
also
transitions
from
pulsations
to
flows
and
back
to
pulsations.
This
shows
that
these
things
are
not
like.
You
know
it's
not
just
that
you
get
pulsations
or
flows,
you
get
both
of
them,
so
you
get
pulses
of
things
and
flows
of
things,
and
this
is
so.
You
can
think
of
flows
as
more
of
a
uniform
movement
through
the
cells,
whereas
pulses
or
these
sudden
things
in
time-
and
this
just
shows
you
what
kind
of
the
transition
between
these
two
mechanisms
across
time
in
different
types
of
tissue,
monolayers,
foreign.
A
A
So
we
suggest
that
the
formation
of
new
cell
cell
contacts
after
washout
resets
individual
cell
polarities
to
generate
long-range
flow
patterns.
So
there's
a
lot
of
you
know,
there's
a
lot
of
artificial
sort
of
manipulation
here.
But
you
can
imagine
that
some
of
these
molecules
could
be
you
know,
there's
sort
of
a
kinetics
of
their
aggregation
and
their
decay
in
the
tissue,
and
so
that
may
play
a
role
as
well.
A
A
In
summary,
of
these
observations
support
the
idea
that
pulsatile
contraction
and
expansion
are
associated
with
variations
in
cell
myosin
intensity
and
that
cells
are
polarized
along
the
direction
of
the
flow.
So
this
is
very
simple
finding
here,
and
so
they
did
some
numerical
simulations.
They
talk
about
that.
They
have
some
examples
here
of
root
pulsations
after
wound,
closure
and
flows
after
washing
up
levastatin,
which
is
a
molecular
thing
in
the
cell,
just
shows
some
of
their
experiments
that
they
observe
so
for
wound
closure.
A
You
have
this
Divergence
and
intensity
curves,
so
the
Divergence
decreases
sort
of
you
have
blue
inclusion.
Then
you
have
pulsation
right
after
one
closure
as
a
wind,
Divergence
of
minimized
and
intensity
maximizes
right
after
Divergence
minimizes,
it's
interesting
Dynamics,
so
I
think
that
the
implication
for
this
is.
We
can
think
about
this
in
terms
of
Networks.
A
So
you
know
these
last
two
papers.
It's
really
kind
of
building
towards
the
theory
of
connectivity
of
the
effects
of
you
know
connective
morphogenesis,
I,
don't
really
know
where
we're
going
necessarily
with
this,
but
I
I
think
there's
some.
There
are
a
lot
of
lessons
that
we
learned
a
lot
of
things
to
think
about,
especially
in
some
of
the
work
we've
been
doing
with
tensegrity
networks
and
embryo
networks,
and
even
maybe
some
of
the
things
with
with
connectoms.
A
So
this
last
paper
is
robust
cell
identity
specifications
through
transitions
in
the
collective
state
of
growing
developmental
systems
and
I'm.
Only
going
to
briefly
talk
about
this
paper
am
I
going
to
go
deeply
into
it.
When
I
think
this
movie
ties
together.
Some
of
the
paper
really
the
first
two
papers
that
we
talked
about.
A
We
review
here
the
main
dynamical
mechanisms
through
which
such
transitions
are
conceptualized
discussed
that
the
differentiation
timing,
robust
cell
type
proportions
and
Recovery
upon
perturbation,
are
emerging
properties
of
proliferating
and
communicating
cell
populations.
We
argue
that
studying
developmental
systems
using
transitions
in
Collective
system,
States
is
necessary
to
describe,
observed,
experimental
factors
and
propose
additionally,
the
basis
of
a
novel
analytical
method
to
deduce
the
relationship
between
single
cell
Dynamics
and
the
collective
symmetry
broken
States
in
cellular
populations.
A
So
they're
arguing
that
in
populations
you
get
the
Symmetry
breaking,
which
you
don't
observe
in
single
cells
they
may
be.
You
know
you
see
differentiation
in
single
cells,
but
you
don't
see
the
collective
state
emerge
from
like
all
single
cells
being
coordinated,
and
you
know
there
being
a
direct
relationship
between
the
cell
and
the
tissue.
A
How
these
specific
regulatory
features
emerged
during
early
development
from
the
Dynamics
of
the
core
Gene
regulatory
Network
in
single
cells
is
therefore
been
a
subject
of
numerous
experimental
and
theoretical
studies
and
they're
cited
right
here.
So
there's
been
a
lot
of
work
on
grns,
a
modeling
grns
and
how
they
can
be
used
to
model
single
cells,
and
then
you
have
these
cell
collectives,
which
actually
we
operate
at
a
different
level
and
of
course,
grns
don't
take
into
account
the
physics
they
just
take
into
account.
A
Probing
the
Dynamics
of
grn's,
underlying
specific
developmental
transitions,
determining
the
exact
intracellular
motifs
and
identifying
the
molecules
and
Implement
them
has
been
there
for
of
significant
interest
to
the
field.
But,
however,
the
mechanism
of
timing,
diverse
developmental
events
has
been
investigated,
both
experimentally
and
theoretically,
and
the
proposed
timers
generally
rely
on
the
dynamical
features
of
grns
and
single
cells.
A
So,
in
this
case,
they're
proposing
a
different
conceptual
framework
by
treating
the
communicating
cellular
population
as
a
single
dynamical
system,
such
as
the
different
cell
types,
that
emerge
collectively
as
a
single
attractor,
so
they're
using
a
tractor
they're
using
dynamical
systems.
Theory
they're,
using
the
concept
of
an
attractor
and
they're
doing
analytical
calculations,
they're
not
really
relying
on
grns
using
communication
between
cells,
and
so
this
is
a
different
approach
than
what
people
are
doing
Gras.
A
Moreover,
in
this
manner
cell
types
emerge
collectively
with
a
precise
timing
and
are
dynamically
maintained
by
cell
cell
communication.
So
this
is
a
you
know.
This
is
kind
of
sets
up
a
distinction
between
grn
models,
which
account
very
well
for
things
going
on
inside
the
cell
and
things
going
on
between
cells,
which
you
know
you
can
use
the
upward
States
of
trns.
A
But
this
is
actually
maybe
a
better
way
is
to
think
about
it
in
terms
of
cell
cell
communication,
so
they're
going
to
talk
about
Landscapes
and
how
cells
go
across
this
landscape
as
they
transition
in
terms
of
their
their
Fates
or
their
identities,
and
the
shows
like
the
potential
for
in
all
cells.
This
is
kind
of
a
theoretical
model.
We
usually
use
grns
to
say
that
once
cells
cross,
this
barrier
this
threshold
to
get
to
a
new
state,
and
we
can
show
that
mechanistically.
A
A
In
a
systems
level,
when
social
is
it
going
towards
some
set
of
States?
So
if
you
have
a
precursor
cell
that
goes
through
multiple
stages
of
differentiation
you
want
to
know,
maybe
you
know
what
the
entire
say.
Maybe
potential
landscape
looks
like
for
that?
Not
just
like
having
switches
go
off
so
yeah,
that's
I,
think
that's
that's
kind
of
what
this
paper
is
about.
It
kind
of
goes
through
some
of
the
differences
between
communication
networks,
and
you
know
these
kind
of
grn
networks
or
stringing
grns
together
and
taking
the
output
and
all
that.
A
A
Back
to
the
other
papers,
where
they
talked
about
long
range
order,
and
it's
a
similar
process
here
and
finally,
you
get
these
dynamical
states
and
the
regulation
of
those
States.
Things
are
in
a
stable,
State
and
then
get
kicked
out
and
they
go
through
these
phase
transitions
and
so
forth.
So
that's
a
nice
paper,
so
I
think
that
sums
up
a
lot
of
theoretical
interests.
It
kind
of
dovetails
a
lot
of
theoretical
interests
that
we're
doing
in
the
group
and
I
hope
you
learned
something.
Thank
you.