►
From YouTube: DevoWorm (2022, Meeting 9): Spatial Transcriptomics, Axolotl models, bioengineering and collectives
Description
Modeling DNA topology from cytogenetics data, creating a spatial transcriptomics layer in embryo models, Axolotl modeling. Select resources for Graph Neural Networks (GNNs). Soft materials, bioengineering, human perception, and collective behaviors in cells and worms. Attendees: Susan Crawford-Young, Karan Lohaan, Richard Gordon, Harikrishna Pillai, Namjote, and Bradly Alicea.
A
A
All
right
so
welcome
to
the
meeting.
I've
got
a
bunch
of
things
to
talk
about.
I
thought
I'd
go
through
some
graph
neural
network
stuff
that
I
found.
I
think
that
might
help
people
if
you're
watching
online
or
you're
watching
in
the
meeting.
It
should
clarify
some
things
for
you.
If
you
have
questions,
I
also
put
some
things
in
the
slack
channel
on
on
graph
neural
networks.
There
was
a
nice
primer.
I
think
I
have
that
today.
It's
one
of
the
things.
A
A
So
I
wanted
to
bring
up
some
information
on
that
and
because
he
wanted
to
put
like
transcriptomic
data
on
top
of
microscopy
data,
and
I
told
him
that
that's
very
hard
to
do.
But,
more
importantly,
I
have
to
look
for
the
right
data
set,
and
so
I
don't
think
I
didn't
see
any
spatial
transcriptomics
on
axolotl,
but
they
do
have
a
lot
on
the
embryo.
B
B
B
C
C
A
C
C
A
C
C
C
C
C
B
A
A
B
C
C
B
A
A
C
A
C
E
A
F
A
A
So
there
are
a
number
of
different
tools
on
there
are.
A
lot
of
you
know
like
atlases
that
people
have
created,
but
a
lot
of
it
is
like
for
specific
genes.
There
are
some
maps
for
fate
mapping.
There
are
some
other
types
of
atlases,
but
they're,
not
really
something
that
you
can
map
spatially
to
the
to
the
embryo.
So
what
you
need
is
something
like
a
spatial
transcriptomics
map
and
it
doesn't
seem
to
exist
for
axolotl.
A
They
have
them
for
other
organisms,
I
believe,
but
not
for
axolotl.
So
this
is
a
paper
that
is
a
manuscript.
A
tissue,
mapped
axolotl
de
novo
transcriptome
enables
identification
of
limb,
regeneration
factors,
and
so
this
is
an
example
of
the
type
of
data
that
they
have
for
this
type
of
work
and
they
have
the
sequence
data
here.
So
there's
a
de
novo
assembly
for
ambi
ambiostoma
mexicanum,
which
is
the
axolotl
the
model
organism
axolotl.
A
A
A
Yeah
yeah,
so
they
have
a
number
of
different
matrices
of
isoforms
and
genes.
So
isoforms
are
different
forms
of
a
gene
they're
like
because
they're,
I
guess,
they're,
proteins
or
rna
that
have
different
sequences.
They
hit
take
on
different
forms
as
they're
when
they're
in
the
cytoplasm,
so
they
can
actually
look
at
the
different
isoforms
by
their
sequence.
A
A
It's
a
map
to
different
tissues,
so
you
can
see
that
they
have
these
different
genes
here
listed
or
these
different
loci
that
they
sequence
and
then
they're
able
to
get
an
expression
level
for
each
of
these.
So
you
can
see
that
they
have
different
numbers
here
when
you
go
over
it
and
you
have
it
for
bone.
Cartilage,
ovary,
testes
blood,
blastoma,.
A
A
So
it's
not
really
specific
to
cells
and
it's
not
really
specific
to
the
embryo.
But
this
is
the
way
they
usually
do
this
type
of
study,
to
look
across
different
tissue
types
and
do
like
a
an
aggregate
sample
of
the
data,
and
then
you
know
for
each
of
these
loci.
So
this
is
where
you
have
these
different
genes
here
that
are
identified
in
different,
so
they're.
You
know
different
types
of
genes
associated
with
human
and
chick
and
mouse
that
come
up.
You
know
if
you
probe
for
it,
you
can
get
a
result.
A
So
this
is
actually
the
tissue
enriched.
I
think
this
may
have
been
a
control,
and
then
this
is
the
actual
tissue
enriched
sample.
You'd
have
to
go
back
to
the
paper
to
read
the
the
methods
of
how
they
did
this,
so
I'm
not
going
to
go
through
those
methods
right
now,
I'm
just
trying
to
give
people
a
feel
for
the
data,
and
then
you
get
these
expression
values
here.
A
A
Other
than
that,
you
know,
there's
there's
some.
You
know
there
are
a
lot
of
papers
where
they
have
like
next
gen
generation
sequencing,
which
is
where
you
get
the
genome.
Plus
you
also
get.
You
know
you
get
the
reads
from
the
genome
sequence
plus
you
also
get
the
expression
values
for
those
reads.
The
reads
are
just
sequences
of
dna
that
they
sample
from
different
parts
of
the
genome
and
the
ones
that
have
high
expression
levels.
A
Then
they
map
it
to
you
know
an
original
draft
of
the
genome
to
see
where
it
is
in
the
genome.
So
what
gene
is
it
associated
with
and
that
sort
of
thing
in
terms
of
getting
it
onto
a
like
some
sort
of
atlas?
It's
it's
a
lot
harder
to
do,
because
you
know
you
can't
you
don't
have
tissue,
you
don't
have
cell
specificity
here
you
just
have
tissue
type
specificity,
and
so
even
here
it's
not
necessarily
going
to
tell
you
exactly
what's
being
expressed
at
some
point
in
development.
A
So
it's
it's
it's
kind
of
a
hard
problem
to
put
all
the
data
together.
I
know
that
I
know
it.
You
know
it
seems
kind
of
easy
to
do
these
things,
especially
with
genomic
data,
because
you
have
to
find
the
right
data
set
for
this
and
then
put
stitch
it
together
with
some
other
data
set
that
involves
microscopy.
A
I
don't
know
of
any
study
where
they've
actually
done
the
time
lapse,
where
they
have
the
embryo
and
they're
following
it
through
time
and
then
they're
getting
like
some
sort
of
you
know,
fluorescent
signal
or
something
like
that
or
where
they
actually
take
different
stages
and
they
sacrifice
them
and
get
the
genomic
data
from
that.
So
that's
that's.
That's
kind
of
the
hard
problem
of
this
spatial
transcriptomics
and
you
know
so,
there's
it
there's.
No,
they
do
spatial
transcriptomics
on
some
organisms,
they've
done
it.
A
A
The
only
way
to
do
this
in
that
axolotl
is
to
do
a
sort
of
a
pseudo
spatial
transcriptomics,
where
you
put
together
like
the
different
tissue
types
or
you
know
the
different
regions.
I
don't
even
think
you
get
for
really
early
embryos.
Given
the
data,
but
you
know
it's
going
to
look
a
lot
different
than
that,
but.
A
B
A
B
A
So
if
yeah,
that
sounds
great,
so
if
you're
interested
in
the
sort
of
spatial
transcriptomic
stuff,
like
some
of
the
stuff
that
people
have
done
so
far,
they
have
papers
on
this.
People
have
generated
some
atlases,
it's
almost
always
on
adult
animals,
but
there's
some
a
little
bit
on
embryos,
but
the
embryos
they're
really
mapped
to
different
cells.
A
So
I
can
really
show
you
a
good
example
of
that
and
then
they
just
you
know
it's
just
a
matter
of
mapping
it
to
some
region
of
the
embryo
or
the
organism,
and
the
hypothesis
is
that
different
cells
are
expressed
differentially
in
different
tissues
or
in
different
areas
of
the
organism,
and
that
tells
us
something
about
you
know
its
its
function.
So
you
know
some
genes,
we,
you
know,
we've
struggled.
We
know
the
sequences
of
a
lot
of
genes,
but
we
don't
necessarily
know
what
their
function
is
and
people
have
done
descriptive
annotations.
A
But
you
know
those
descriptive
annotations
are
usually
pretty
poor
to
really
get
a
good
handle
on
function.
It's
usually
something
like
cell
maintenance
or
you
know.
If
it's
there's
some
strong
effect
of
the
gene,
they
know
kind
of
what
it
is,
but
it
you
know
breaking
it
out
by
space
is
a
lot
easier
to
say.
Well,
okay,
that's
obviously
involved
in
something
in
the
head.
You
know
it's,
so
it's
it's
still
a
lot,
there's
a
lot
of
problems
with
it,
but
it's
it
improves
our
progressivity.
C
A
A
A
A
F
A
H
A
A
G
A
H
F
H
H
Yeah,
okay,
I
did,
I
think,
zero
and
on
some
key
issues.
You
know
that
could
help
me
improve
the
model
to
you
know
because
3d
reconstruction
doing
it
in
like
requires
some
special
techniques.
You
know
like
you
need
camera
parameters.
You
know
focal
length
distance
from
the
object
and
iso
format
and
all
those
things
the
ones
that
are
you
know
exist.
H
C
C
Okay,
you
want
me
to
do
it
now.
Yeah
go
ahead!
Oh
okay!
Okay!
Let
me,
let's
see
you
can
see
me
so
I'm
going
to
hold
up
a
book.
That's
the
quickest
way
to
find
a
picture.
Okay!
This
is
a
an
aberrant
diatom.
That
means
it's
it's
screw
up
in
terms
of
its
shape
and
morphing
fences,
here's
another
one:
okay
and
here's
another
one
they're
all
the
same
species,
and
I
have
about
a
hundred
electron,
micrographs
of
them
from
ryan
drum.
C
C
Yeah,
okay-
in
other
words
you
have
to
you,
have
to
find
all
the
lines,
I
guess
and
find
their
width
and
put
them
and
see
if
they
form
a
bimodal
class
and
then
isolate
all
little
lines
that
are
thick
yeah,
which
is
obviously
not
a
trivial
image.
Processing
problem:
okay,
okay
and
the
thick
lines
can
be
broken
into
a
few
parts,
yeah
okay,
so
the
idea
is
to
use
the
thick
lines
as
a
start.
C
C
B
C
B
Put
it
on
into
my
computer
and
my
computer
crashes,
because
it
says
it
runs
out
of
memory
and
I
have
16
gigabytes
of
memory
on
my
computer,
so
they
need
to
be
more
of
a
sparse
set
mind
you.
I
was
going
to
try
to
give
that
to
dr
sharif,
my
advisor
for
one
of
his
students
or
something,
but
I
mean
I'm
just
saying
I
have
this
data
and
tom
portages
worked
on
it
already
and
got
it
into
tiff
files
from
for
that
matter,
as
slices
anyways
are
these
projection
data
yeah?
B
C
B
C
B
B
C
C
B
J
B
H
A
E
F
J
A
Looks
pretty
good,
so
that's
this!
That's
where
you
can
turn
it
with
your
cursor
and
yeah.
It
looks
like
looks
like
it's
pretty
good.
You
know,
there's
some
yeah
look
at
there.
It's
there's
a
little
bit
of
like
a
zone
where
it's
sort
of
stitched
together,
but
that's
good.
I
mean
that's,
probably
yeah.
I
don't
know
if
you
are.
J
A
That
looks
good
yeah.
I
I
enjoy
that
you
know
I
had
mentioned
like
you
know.
One
way
to
do
this
is
to
take
the
images
and
put
them
on
to
some
sort
of
sphere
and
map
them
kind
of
like
warp
them
from
the
flat
surface
to
some
sort
of
spherical
representation,
and
then
you
know
something
that
you
can
manipulate
in
a
browser,
so
that
was
made
with
this
electron
program.
Correct.
A
J
A
A
That's
very
good
now
I'd
like
to
talk
about
a
couple
things,
I'd
like
to
talk
about
graph,
neural
networks
and
then
go
on
to
a
couple
other
items
and
then
our
paper
of
the
of
the
week.
So
this
is
something
I
posted
in
the
in
the
slack:
the
open
worm,
slack
and
the
diva
worm
channel.
This
is
graph
neural
networks
for
novice
math
fanatics,
and
this
is
something
that
is
created
by
rashabanand.
A
So
this
was
something
that,
and
you
can
see
that
there
are
a
number
of
different
he's
drawn
out
a
lot
of
examples
here.
So
this
is
a
nice
thing
that
you
can
follow
if
you're
interested,
so
basically
just
kind
of
goes
over
this
distill
article
that
he
posted
before
so
distill,
of
course,
is
this
journal
that
has
a
lot
of
machine
learning
articles
on
it
and
it
has
a
lot
of
interactive
articles.
A
It's
a
very
nice
pla,
publishing
platform
they've
taken
a
break
right
now,
but
you
know
there
are
a
number
of
different
groups
that
are
doing
these
articles
so
this.
Actually,
this
article
is
a
general
introduction
to
graph
neural
networks,
and
it
just
shows
you
like
you
know
how
these
these
embeddings
are
created
and
work.
A
So
you
can
touch
these
different
layers
and
they
show
you
how
the
properties
of
these
different
layers
with
respect
to
the
graph
embeddings
in
each
layer.
So
you
can
see
that
things
propagate
out
like
they
would
in
a
neural
network,
but
each
layer
has
a
graph
associated
with
it
so
graph
embedding
where
the
nodes
are.
A
A
So
if
you're
kind
of
stuck
on
this
graph
neural
networks
concept,
this
is
a
good
way
to
kind
of
get
around
those
those
barriers,
and
so
you
know
we're
representing
graphs.
We
basically
use
this
adjacency
matrix
and
we
have
a
it's
a
binary
representation
here
where
we
want
to
know
you
know
if
we
want
to
look
at
these
nodes
one
through
six.
How
are
they
connected
together,
so
they
can
be
connected
to
themselves
or
connected
to
any
of
their
other
neighbors,
and
so
this
adjacency
matrix
tells
us
what
those
connections
look
like.
A
If
it's
an
all
to
all
connectivity,
all
of
these
values
would
be
one,
so
everything
would
be
one
every
cell
would
be
connected
to
itself.
Every
cell
would
be
connected
to
its
neighbor
or
any
other
cell
in
the
in
the
group.
You
know
this
is
kind
of
like
a
markov
model
where
you
have
cells
that
are
connected
to
themselves
and
then
they're
connected
directionally
to
some
neighbor.
So
three
is
connected
to
six,
but
of
course,
six
isn't
connected
or
actually
I
think
six
is
connected
to
three.
So
this
is
a
bi-directional
graph,
but.
A
And
so
in
that
way,
it's
like
a
markov
model
and
it
you
know
the
advantage
using
a
complex
network.
Of
course.
Is
you
get
these
parallel
representations
of
different
variables
and
that
they're
connected
you
know
to
different
variables
in
different
ways,
so
you're
able
to
get
these
higher
order
structures
like
cliques
and
modules,
which
are
just
like
local
groups
of
cells
that
are
connected
to
one
another?
A
And
so
that's
that's.
Basically
you
build
a
graph.
Then
you
have
these
and
of
course,
this
type
of
graph
is
unweighted,
so
it's
zeros
and
ones.
But
if
you
had
a
weighted
graph,
which
you
have
in
a
lot
of
neural
networks,
not
on
a
graph
but
on
a
on
a
network
of
cells
that
are
connected,
you
have
this
decimal
value
between
zero
and
one.
A
So
it's
you
know,
that's
that's
the
way
you
build
graphs
and
then
this
is
what
a
complex
graph
looks
like
when
you
have
a
lot
of
nodes
or
cells,
and
it
has
this
sort
of
what
they
call
hairball
quality.
So
the
hairball
quality
is
where
it
has
this
sort
of
highly
connected
topology
and
it
kind
of
looks
like
a
hairball
in
a
sink.
It
has
this.
A
So
this
is
an
image,
and
then
this
is
the
graph
that
this
image
you
could.
You
can
create
a
graph
to
represent
this
image,
and
so
this
is
the
kind
of
thing
we
want
to
do.
We
want
to
take
like
images.
In
our
case,
we
want
to
take
a
subset
of
that
image.
We
want
to
segment
the
image
in
some
way.
It
doesn't
have
to
be
like
you
know,
traditional
image,
segmentation.
A
A
So
you
know
there's
a
lot
more
here
in
terms
of
like
looking
at
edges
and
explaining
some
of
the
details
here,
message
passing
that
won't
necessarily
be
something
that
you'll
need
to
worry
about
for
your
projects,
but
that's
something
we
should
talk
about
in
the
future.
There's
this
other
thread
about.
So
this
is
the
one
I
posted
in
the
slack
on
graph
theory
and
gnns
can
be
scary
at
first
with
so
many
architectures.
A
So
this
is
a
maze
analogy,
that's
used
here
and
this
kind
of
talks
about
the
top
six
strategies
for
navigating
a
maze
and
this
talks
about.
You
know
how:
how
network,
how
you
navigate
a
network
and
they're
alternate
strategies
for
this
navigation,
and
so
it
kind
of
goes
through
this.
You
know
kind
of
talks
about
how
you
would
build
network
embedding
so
there
you
know
you
can
go
through
this
article,
and
this
is
a
nice
set
of
visual
aids
that
you
know
it
tells
you
how
to
go
through
the
network.
A
A
For
novice,
so
this
is
the
one
for
novices
by
rashav
anand.
This
one
is
the
distilled
article.
This
one
is
the
publishing
published
in
towards
okay.
This
is
up
towards
data
science.
This
is
graph
neural
networks,
and
this
talks
about
message
passing
and
physics
inspired,
continuous
learning
models
and
graphs.
This
is
a
very
specific
example
of
the
sort
of
graph
ml,
so
they
uses
the
example
message
passing
paradigms
where
people
are
learning
on
graphs,
so
in
the
graph
community.
A
He
argues
that
the
node
and
edge-centric
mindset
of
current
graph
deep
learning
schemes
imposes
unsurmountable
limitations
that
obstruct
future
progress
in
the
field.
As
an
alternative,
I
propose
physics-inspired
continuous
learning
models,
so
these
are
just
a
different
type
of
model
for
doing
this.
For
looking
at
how
these
graphs
work
and
their
potential,
it
opened
up
a
new
trove
of
tools
for
the
fields
of
differential
geometry,
algebraic,
topology
and
differential
equations.
A
So
far
large
a
lot
explored
in
graph
ml.
So,
if
you're
interested
in
the
potential
of
this
in
other
areas,
this
is
something
you
want
to
read.
I
can
put
this
in
the
slack
as
well,
but
it
does
go
through
some
of
these
basics
here
at
the
beginning
and
so
also
talks
about
message
passing
and
things
like
that.
A
Again,
those
aren't
things
you
need
to
know
for
the
project,
but
those
are
useful
things
to
know
and
so
again
the
graphs
graph
motifs
now
so
another
I
mentioned
cliques
and
I
mentioned
modules,
but
the
basic
unit
of
a
graph
is
a
motif
and
so
beyond,
like
the
nodes
and
the
edges.
So
a
motif
is
like
where
you
have
this
relationship
between
x
and
y,
and
you
have
like
these
little
things
like
triangles
or
zigzags,
or
things
like
that
that
you
can
identify
again
and
again
in
the
network.
So
you
saw
that
hairball
before
that.
A
Hairball
can
be
broken
down
into
these
motifs.
It
can
also
be
broken
down
in
other
kinds
of
structures,
but
the
motifs
are
very
easy
interesting
because
they
describe
certain
relationships,
so
very
simple
relationships
like
transcriptional
networks.
How
gene
x
is
affecting
gene
y
or
a
neuron
synaptic
connection
network.
How
neuron
x
is
affecting
neuron
y
or
even
a
food
web
where
predator
acts
as
a
affecting
prey
y,
and
so
that
kind
of
puts
maybe
not
only
allows
you
to
find
different
patterns
in
the
graph
to
see.
A
If
you
know
these
things
repeat
or
if
they're
you
know
in
certain
parts
of
the
graph,
but
also
allows
you
to
figure,
maybe
a
figure
out
causality,
and
so
that's
something
that
again,
you
don't
need
to
know
for
this
project,
but
it's
useful
and
then
they
have
these
interpretable
gnn
models,
which
is
where
you
have.
You
know,
you're
looking
say
at
the
chemical
structure
of
something
it
gets
put
on
a
surface.
A
A
A
So
this
is
our
hypergraph,
where
you
have
the
single
cell
and
you
get
these
different
regions
of
the
embryo,
these
different
cell
types
and
they
form
different
sub
networks
and
then
there's
their
connections
between
these
sub
networks
shown
in
the
graph.
And
so
the
idea
is
that
these
hyper
graphs,
each
of
these
nodes,
have
a
sub
graph
within
them.
A
Like
you
see
here
some
of
these
cells,
you
know
they
get
distributed
from
a
single
undifferentiated
network
into
sort
of
like
a
generic
network
and
in
a
connectome
you
know,
so
you
see
that
those
sort
of
relationships,
and
so
that
that's
you
know
I
I'll.
I
want
to
work
in
this
into
a
full
paper.
So
next
question
yeah.
Is
there.
A
Yeah
yeah
so
yeah
that
that's
that
I'd
like
to
work
this
into
a
full
paper,
and
if
people
are
interested
in
working
on
it
with
me,
that
would
be
great.
We
could
work
on
a
you
know,
maybe
something
where
we
implement
something
that
is
more
empirically
relevant.
So
I
also
have
found
these
things
on
my
on
the
synthetic
daisies
blog
a
couple
of
articles
on
soft
materials,
I
thought
would
be
interesting,
so
there's
this
merging
electronics
in
biology
the
future
of
touch.
A
You
know
you
can
tell
that
there
are
differences
in
those
surfaces,
and
so
the
surface
properties
affect
how
you
touch
things
and
how
you
perceive
you
know
your
movements
against
those
materials.
So
if
you
walk
on
something
it's
very
soft
versus
very
hard,
you
can
tell
if
you're
touching,
against
something
pressing
in
something
that's
very
hard
versus
very
soft.
A
You
can
also
tell-
and
there
are
different
consequences
to
that
and,
like
you
know,
in
terms
of
sensation
and
muscle
activation,
and
things
like
that,
this
is
actually
then
they're
like
I
used
to
do
these
posts,
where
I
would
like
go
across
the
number
of
different
topics,
so
this
kind
of
gets
into
something
called
conformal
electronics,
which
is
a
nice
way
that
they
use
compliant
materials.
Like
you
know,
pieces
of
plastic,
but
they
embed.
A
J
A
But
you
know
some
of
them
have
penetrated
the
commercial
market
by
now,
and
so
you
know
it's
like
where
you
have
this
piece
of
plastic
that
bends,
but
you
can
also
project
things
onto
it
like
you
can
use
it
as
an
e-reader
or
you
know
you
could
use
it
as
some
sort
of
computer
display.
So
that's
something
that
you
know
it's
kind
of
has
implications.
A
A
A
So
they've
made
it
from
muscle
cells
and
it
has
a.
I
think
it
has
a
pacemaker
in
it
and
then
it
basically
it
pulses
so
that
it's
able
to
propel
itself
through
the
water
they
call
it
a
medusoid
and
the
thing
swims
through
the
water
column
and
because
the
video
will
construct
this
thing.
They
kind
of
know
what
it's
going
to
do,
so
it's
this
muscular
pump
that
they've
built
they're
building
it
in
this,
like
as
kind
of
like
a
jellyfish,
it's
a
bio-inspired
device.
A
It
looks
like
a
jellyfish,
but
they
call
it
a
medusoid,
and
so
this
is
a
stripped
down
version
of
the
joey
fish
morphology
replicating
only
the
components
needed
to
approximate
jellyfish
swimming.
So
that's
all
that
you
need
in
this
jellyfish
jellyfish
aren't
that
complex,
but
they
do
have
a
they
do,
have
what
they
call
a
nerve
net,
so
they
propagate
signals
throughout
their
body.
A
So
these
are
the
cells,
the
precursor
cells
in
in
the
heart
that
contribute
to
formation
of
a
heart
muscle
and
things
like
that.
We're
allowed
to
self-assemble
into
the
desired
structure,
so
they're,
building
these
out
of
rack
cardiomyocytes,
but
they're
they
resemble
jellyfish
cardiomyocytes,
will
spontaneously
contract
in
culture
which
enables
the
cell
population
to
approximate
a
nerve
net.
So
this
is,
I
told
you
about
these
nerve
nets
and
they're,
just
basically
where
you
have
the
nervous
system
distributed
throughout
the
morphology.
A
A
A
They
differentiate
into
some
sort
of
excitable
muscle
cell
and
they
can
do
the
same
thing
if
they're
on
this,
you
know
they're
not
forming
the
heart
anymore
they're
forming
this
structure,
and
so
you
can
see
that
there's
this
pacemaker
system
around
the
edges
of
the
jellyfish
and
you
can
do
the
same
thing
in
a
medusa
with
cardiomyocyte,
and
you
can
stimulate
it
in
a
similar
way.
Then
they
show
the
stroke,
kinematics
and
what's
happening.
A
A
They
were
able
to
do
some.
They
were
able
to
do
some
gfp
on
this
to
show
some
of
the
details
of
the
muscle
fibers
here
and
the
medusoid
muscle
and
the
jellyfish
muscle,
so
the
jellyfish
muscle,
medusoid
muscle-
I
mean
you
know
their
muscle
fibers,
but
they
just
wanted
to
show
the
difference
between
them.
A
A
So
you
know
it
does
behave
somewhat
like
jellyfish
and
then
there's
some
like
goodies
at
the
end
here,
just
kind
of
things.
This
is
like
gonna
like
putting
it
into
some
context,
yeah,
so
there's
their
cardiac
pacemaker
cells.
Those
are
the
cells
that
allow
the
muscle
to
keep
beating
and
they're
they've
done
a
lot
of
simulation
of
these
neurons,
but
there's
also
muscle
that
gets
that
does
a
lot
of
the
work.
A
A
A
You
know
different
types
of
machines
that
resemble
things
in
biology
so
and
then
we've
also
talked
about
how
matter
makes
a
difference
in
perception
and
human
perception
when
it's
touched
and
when
it's
interacted
with
this
is
going
more
in
the
direction
of
kind
of
like
this
biological
engineering.
So
this
paper
is
collective
motion
of
cells
modeled
as
ring
polymers
and
so
in
this,
and
I
think
susan
sent
this
to
me
or
dick.
I
can't
remember
who-
but
this
is
from
the
soft
matter
journal,
so
this
is
this
paper.
A
A
As
a
result,
though,
there's
this
collective
behavior
that
where
cells
are
pushing
against
one
another,
because
they
have
neighbors
in
all
directions,
usually
when
they're
in
culture,
sometimes
if
a
culture
is
growing,
it's
expanding
outward,
but
still
you
have
cells
in
the
in
interior
of
that
mass.
That
are,
you
know
having
to
that,
are
subject
to
some
of
these
competitive
forces
of
staying
away
from
its
neighbor.
A
So
this
cooperative
motion
emerges
as
the
amplitude
of
the
motility
force
as
the
amplitude
of
motility
forces
increased
and
where
their
aerial
density
is
increased.
The
degree
of
collectivity
characterized
by
the
average
cluster
size,
the
velocity
field,
order,
parameter
and
polarity
field,
pneumatic
order
parameter
so
pneumatic.
We
talked
about
that
and
they're
using
those
different
terms.
You
know.
A
So
then,
furthermore,
the
degree
of
alignment
exhibited
by
the
cell
velocity
field
within
a
cluster
is
found
to
be
stronger
than
the
exhibited
by
the
cell
than
that
exhibited
by
the
cell
polarity
comparison
between
the
collective
behavior
of
elongated
cells
and
that
of
circular
cells.
So
cells
have
different
morphologies.
Some
are
elongated
along
their
axis
and
some
are
circular,
so
they're
just
circles,
and
this
gives
you
different
behaviors
and
different
collective
behaviors.
A
G
C
A
A
C
A
Right,
so
you
can
see
that
you
have
these.
These
are
the
cells.
These
are
elongated
cells
and
then
b
is
the.
Let's
see
a
is
sort
of
the
snapshot
of
cells
b
is
a
snapshot
of
cell
velocity.
So
these
arrows
are
the
cell
velocities
as
they're
moving
against
one
another,
and
then
c
is
the
snapshot
of
cell
polarities
of
the
same
system
shown
in
a
so.
These
lines
are
like
how
they're
oriented
if
they're
oriented
you
know
northeast
southwest
or
north
south
or
east
west.
A
I'm
just
using
directions
to
describe
that,
but
you
can
see
that
there's
like
some
collective
behavior
there
somewhere
there's
a
synchronization
and
in
other
cases
where
it's
misaligned
relative
to
its
neighbors.
But
you
can
see
that
you
get
these
emergent
structures
where
they're
all
kind
of
aligned
in
the
same
way
in
certain
places.
A
And
so
this
is
just
where
the
cells
are
on
this
on
the
surface
they're
allowed
to
grow
out,
and
then
they
show
these
polarity
or
these
velocity
arrow,
like
they
just
show
the
arrow
and
it
shows
the
direction
of
movement,
and
I
guess
the
velocity
field,
which
is
also
the
length
of
movement.
So
I
think
these
arrows
are
of
different
lengths
a
little
bit.
A
So
that's
that's
what
we
have
here
and
you
can
see
that
they
form
and
if
you
look
at
it,
it
just
looks
like
they're
forming
random
patterns,
but
the
there's
a
sorting
that
goes
on
as
they're
growing
and
kind
of
growing
randomly
that
they
start
start
to
form
these
clusters
and
they
move
against
one
another,
and
then
they
form
these
colonies.
So
these
colonies
look
like
this,
so
the
other
article
here
is
this.
A
This
is
from
also
from
the
soft
matter
journal,
and
this
is
synchronized
oscillations
and
swarms
of
nematode.
So
this
is
something
again
with
nematodes
which
we're
familiar
with
the
c
elegans.
Well,
this
is
a
different
species
in
order
or
genus
of
nematode
turbatrix
aceti.
So
this
is
turbatrix
as
the
genus,
and
these
are
just
synchronized
oscillations
in
these
swarms.
So
nematodes
can
form
these
collective
balls
of
of
nematodes
in
certain
environments
and
you
look
under
a
microscope.
Sometimes
you
can
see
them
all
sort
of
congealed
into
certain
these
clusters.
G
A
G
A
So
there's
been
a
recent
surge
of
interest
in
the
behavior
of
active
particles
that
can
both
align
their
direction
and
movement
and
synchronize
their
oscillations
and
those
are
called
swarm
elators.
So
that's
interesting.
I
didn't
know
that.
Well,
theoretical
and
numerical
models
of
such
systems
are
now
abundant.
Now
real
life
examples
have
been
shown
to
date,
so
this
is
an
example
showing
the
collective
motion
of
this
nematode.
A
That's
self-propelled
with
body
undulation,
so
I
don't
know
some
p.
The
newer
people
have
seen
the
c
elegans
motive
movement,
but
they
do
this
sort
of
side
winding
and
it
or
not,
sidewinding,
but
it
kind
of
like
going
back
and
forth
and
you
know
like
they're
swimming,
and
that
is
body
undulation.
There
are
different
ways
that
they
can
undulate
their
body,
it's
basically
that
they
curve
and
they
move
forward
and
they
have
different
modes
of
movement.
If
you
want
to
see
the
different
modes
of
movement,
you
can
look
it
up
they're.
A
You
know
different
observations
of
this
there's
like
a
main
mode
and
then
there's
like
a
swimming
in
liquid
mode.
So
there
are
different
ways
that
they
move.
We
discover
that
these
nematodes
can
synchronize
their
body
oscillations
forming
striking
traveling
metacronal
waves.
So
I
guess
these
are
time
dependent
waves
which
produce
strong
fluid
flows,
so
they
have
like
a
wake
and
they
generate
a
flow.
We
uncover
that
the
location
and
strength
of
this
collective
state
can
be
controlled
through
the
shape
of
a
confining
structure,
in
this
case
the
contact
angle
of
droplet.
A
So
this
is,
let's
see
so
this
is
a
this
photos
of
evaporation
of
a
250
microliter
droplet.
So
this
is
the
evaporation
of
the
droplet.
Here
you
have
this
initial
density
of
the.
So
as
this
droplet
is
evaporating,
these
nematodes
sort
of
congeal
here
in
this
ball
in
this
swarm,
and
then
here's
here,
the
nematodes
moving
against
one
another
in
this.
In
this
ball
they're
synchronized
movements
here,
so
you
can
see
that
they're
all
kind
of
aligned
and
they
form
this
metacronal
wave.
A
So
you
can
see
this
metacronal
wave
under
a
microscope,
they're
all
synchronized
in
time,
so
they
have
this.
They
have
these
movements,
here's
the
undulating
movement
where
they're
kind
of
curved
and
they
curve
back
and
forth.
So
this
part
moves
up
and
this
part
moves
down
and
they
undulate
in
that
way
and.
A
C
A
They
have
like
they,
they
get
next
to
one
another,
but
there's
they
don't
do
this
sort
of
thing
they
kind
of
like
they
kind
of
go,
move
against
one
another
back
and
forth,
and
they
kind
of
that.
The
males
are
there's
a
very
small
population
of
males
that
have
a
they
have
like
a
spike
on
their
tail
and
that's
how
they
mate
with
the
hermaphrodites
and
they
you
know
if
they.
A
Back
and
forth
against
one
another
actually
in
this
way,
they
don't
do
this
kind
of
like
organization
where
they're
creating
you
know
they're
moving
against
one
another
like
this,
where
they're
creating
a
wave,
it's
just
kind
of
like
sliding
against
one
another
kind
of
interesting,
but
so
yeah.
This
is,
and
so
you
know,
nematodes
can
form
a
lot
of
these
kind
of
collective
behaviors.
A
A
You
know
when
it's
desiccating,
but
you
also
have
this
experimental
model
where
they're
allowing
they
have
the
the
bunch
of
nematodes
in
a
droplet,
because
the
nematodes
are
very
small,
they
put
the
droplet
down
and
then
it
evaporates
and
then,
as
there
are
a
lot
of
nematodes
in
this
droplet,
you
can
suspend
the
nematodes
in
like
a
solution
and
put
down
a
droplet
and
there's
so
many
nematodes
in
it
and
then,
as
the
droplet
evaporates,
they
kind
of
come
together
and
form
this
collective.
So
but
then
they
explore
this
on.
A
A
You
know
it's
not
a
very
big
force,
but
there
is
a
force,
and
especially
collectively
you
know
at
the
scale
the
size
scale,
they're,
creating
a
pretty
decent
size
force,
and
so
they
show
the
contact
angle
of
the
droplet
and
then
the
force
of
the
nematodes
against
that
all
right.
So
I
think
that's
enough
for
today
on
that,
so.