►
From YouTube: DevoWorm #32: Open Source, Symmetry-breaking and movement, and simulating Drosophila eye development
Description
Open Source contributions. Symmetry-breaking as demonstrated by cell motility and biophysical dynamics. Analysis and simulation of Drosophila eye development with high-resolution cell segmentation and EYEvolve. Attendees: Alon Samuel, Richard Gordon, Bradly Alicea, Susan Crawford-Young, Morgan Hough, Anant Kumar, and Karan Lohaan
B
A
B
A
C
I
didn't
do
much
afterwards,
so
I
don't
have
much
to
update
beyond
it
yeah
and
there
is
a
bit
of
stuff
that
I
saw
in
the
in
the
repos
that
I
didn't
get.
So
what
tarou
made
like
a
pull
request
that
maybe
I'm
gonna
go
through
later,
and
I
made
a
pull
request
to
another
repo,
the
the
the
notebook,
the
demo
data
science
demo.
A
B
Yeah,
that's
it
all
right,
yeah,
that's
great
yeah,
so
it
was
a
good
meeting.
I
thought
it
was
very
informative
kind
of
got
into
exactly
what
thomas
has
been
doing
with
respect
to
he's
been
collecting
data
on
the
vassal
area,
which
is
the
type
of
diatom
he's
like
he's,
got
his
own
setup
in
his
house.
I
think
where
he
goes
and
gets
diatoms
from
the
local
river
and
he
puts
him
under
there
and
he
takes
movies
and
he's
you
know
it's
not
a
sophisticated
microscope.
B
It's
just
something
that
will
pick
up
that
resolution
and
makes
these
movies,
and
then
he
can
handle
he's
a
former
physicist,
so
he
can
or
he's
a
physicist
now
I
guess,
he's
doing
physics
type
analysis
of
the
video.
So
it's
a
little
bit
different
than
like
the
typical
machine
learning
and
you
know
it'd
be
nice
to
have
like
other
types
of
analysis
applied
to
it
and
yeah.
C
Yeah
yeah,
although
it
didn't,
I
don't
share
much
of
the
results
of
the
synthetic
videos
at
the
top
but
yeah.
It
was
interesting
to
to
think
of
like
from
the
synthetic
videos
what
we
kind
of
what
noise
you
can
see
from
that
then
kind
of
describe
what
kind
of
maybe
the
best
measurements
of
the
movement
that
we
can
get
so
yeah.
A
B
C
A
B
B
D
B
B
We
have
the
graph
neural
networks,
work,
that's
being
done
by
wataru
and
jiahung
and
they're.
You
know
they're,
I
think
they're
at
a
good
point
where
they
can
submit
something
to
gsoc.
So
that's
good
and
then
they'll
pro,
and
at
least
you
hung
will
continue
in
into
the
fall
and
yeah.
It
looks
pretty
good
that
you
know
they
need
to
do
some
more
training
on
data
and
things
like
that.
But
that's
always
a
challenge.
B
Okay
and
then
the
other
project,
digital
microspheres.
We
have
koran
and
hari
krishna
and
their
work
is
both.
You
know
they're
both
on
track
to
submit
they
have,
they
were
trying
to.
I
you
know
I
told
them
don't
but-
and
you
know
don't
let
good
be
the
enemy
of
perfect.
I
guess
is
how
you
say
that
just
you
know,
if
you
have
something
it's
a
proof
of
concept,
that's
that's
good
enough
for
submission.
You
don't
need
to.
B
B
I
look
forward
to
seeing
it
and
so
then
we'll
have
this
three-dimensional
atlas
on
a
sphere
that
we
can
explore
things
different
types
of
embryo
data
with
so
now
we
can
generate
more
data
with
things
like
the
flipping
microscope
and
the
ball
microscope
and
other
types
of
microscopes
and
get
the
data
into
the
system,
and
I
don't
know
what
yeah.
This
is
something
that
we
can
add
on
to
later.
It'll
be
kind
of
like
evil
learn.
We
have
like
a
variable.
We
want
to
measure,
we
can
add
it
in
later.
A
I
could
still
make
those
the
bbs
and
into
I
put
stripes
on
try
putting
stripes
on
the
ball,
bearings
and
bbs
that
I've
got
so
that
I
can
put
them
in
my
ball
microscope
and
then
see
how
well
these
approaches
work.
So
I
was
going
to
try
to
put
stripes
this
way
that
way.
This
way
you
know
just
so,
you
can
can
see
if
it,
if
how
well
that
works,
how
the
lines
align
or
something
on
the
surface.
So
I'll
try
to
do
that.
I
can
do
that
now,
yeah.
B
C
B
A
A
C
C
B
B
C
C
C
B
C
C
C
Matching
algorithms
need
to
be
refined
a
bit
more
further
to
improve
the
the
angle
of
rotation
calculations,
so
that
is
also
required
because
current
optical
flow
or
any
any
other,
you
know,
feature
matching
algorithm
that
is
used
for
tracking
the
rotational
motion
of
any
body.
So
I've
kind
of
outlined
those
things
in
the
future
plan
in
that
report
repository
itself,
okay,
so
yeah!
So.
B
C
C
C
C
Okay,
okay
yeah,
I
was
saying
you
know
I'll
continue
on
keep
on
updating
these
changes.
You
know
that
I've
outlined
the
future
plans
in
the
coming
week
section.
You
know
we
get
a
good
like
we
define
it
to
a
certain
degree.
You
know,
because,
when
we're
dealing
with
the
small
set
of
images,
you
know
the
insight
about
how
accurate
the
final
model
is
compared
to
ground
reality
and
all
those
kind
of
metrics
have
to
be
there.
So
these
are
a
few
things
for
that.
C
Yeah
I
was
planning
on
going
forward
with
that
on
the
erupa.
You
know
trying
the
app
on
heroku
and
showing
the
model
in
an
interactive
using
in
proper
ui,
but
I
was
not
able
to
you
know
like
on
heroku,
I'm
still
facing
some
issues
on
deploying
the
python
scripts,
so
I'll
continue
doing
that.
Otherwise,
the
repository
is
kind
of
updated,
with
the
changes
that
I've
made
so
far.
Okay,
so
kind
of
you
know
filling
out
that
for
the
gsoc
file
submission
thing,
so
I've
included
that
this
this
part
also
included.
C
C
C
Like
a
platform
where
you
know
you
can
deploy
python
very
easily,
for
you
know,
instead
of
going
with
a
full-fledged
website-
and
you
know
dealing
with
all
the
back-end
stuff
using
class
for
django,
they
offer
a
very
you
know
easy
interface
to
deploy
all
sorts
of
python
scripts
to
you
know.
So,
if
you
have
like
good
visualization
tools
inside
your
python
scripts
as
well,
so
you
can,
you
know,
have
a
full-fledged
application
that
can
be
deployed
within
a
few
days.
C
C
B
All
right:
well,
it's
good
thanks
for
the
update
and
yeah,
it's
a
very,
very
active
time
for
the
project
kind
of
getting
to
the
end
of
it.
Getting
everything
like
the
everything
kind
of
the
final
touches
on
it,
so
so
the
I'm
gonna
actually
be
attending
a
session
for
the
incf,
so
the
incf
yeah,
oh
morgan,
you
were
going
to
say
something.
B
B
C
B
B
A
lot
of
standards
for
neural
imaging,
a
lot
of
standards
for
data
analysis
and
data
archiving,
and
things
like
that.
So
their
focus
is
on
sort
of
neuroinformatics
broadly
how
to
take
data
sets
and
work
with
them.
They
have
these
fair
standards,
which
are
you
know
a
way
to
like
a
set
of
standards
for
people
to
distribute
data
to
archive
it
to
make
it
findable,
accessible,
interoperable
and
reproducible.
So
that's
what
fair
stands
for
and
it
goes
through
a
lot
of
these
different
infrastructures.
B
So
I
I
don't
know
if
they
have
the
agenda
for
2022
on
here,
but
I'm
actually
presenting
a
poster
on
some
other
work
that
I've
been
doing
with
another
group
that
I
work
with,
and
you
know
it
should
be
a
good
time.
I
hope
that
you
know.
I
hope
that
they,
you
know
that
there's
some
really
exciting
things
going
on.
I
hope
to
hear
about
the
updates
on
that
and
yeah,
so
I
just
wanted
to
put
a
plug
in
for
that.
B
So
we
have
a
lot
of
things
kind
of
backed
up
here.
Here's
one
thing
that
we
talked
about
in
the
past,
this
push-pull
morphogenesis
stuff.
So
this
is
like
the
fourth
installment
of
this.
I
think
the
last
time
I
talked
about
a
paper
on,
I
can't
remember
what
it
was,
but
this
is
this
paper
is
on
3d
single
cell
migration.
B
This
paper
is
called
3d
single
cell
migration,
driven
by
temporal
correlation
between
oscillating
force
dipoles,
so
it's
a
very
physic
physics,
biophysics
oriented
paper,
and
it's
about
single
cell
migration
and
visualizing
this
in
3d,
and
so
this
is
actually
very
much
in
the
spirit
of
also
what
susan
was
talking
about
with
that
paper
that
she
wants
to
send
me.
So
this
is,
let's
see
not
familiar
with
any
of
the
authors.
B
The
abstract
reads:
directional
cell
locomotion
requires
symmetry
breaking,
which
is
where
you
have
a
symmetrical
structure
and
there's
some
process
that
makes
it
asymmetrical.
Symmetry
breaking
is
a
concept
from
physics,
but
they
use
it
in
development.
A
lot
to
discuss
embryos
that
become
asymmetrical.
D
B
You
know
if
you
have
that
pencil
and
you
let
it
drop,
and
it
goes
one
way
or
another,
it's
sort
of
an
entropic
decision,
so
the
pencil
has
to
you
know
it
can't
levitate
in
the
air.
It
has
to
go
in
one
direction
or
another,
and
so
you
know
it
exhibits
a
sort
of
entropy
where
it
makes
a
decision
and
it
gets
sort
of
put
into
a
certain.
You
know
down
a
certain
path
of
possibilities.
B
It
can't
fall
in
the
other
direction
until
you
reset
it
so
well.
Anyways,
directional
cell
locomotion
requires
symmetry
breaking
between
the
front
and
rear
of
the
cell.
So
this
is
where
you
have
a
single
cell,
that's
circular.
So
if
you
have
a
single
cell
and
I'll
draw
it
out
right
here,
consider
a
single
cell,
the
nucleus,
this
is
the
front,
say
anterior
and
posterior.
B
So
sometimes
it
gets
these
protrusions
that
come
out
and
the
idea
is
they
make
these
protrusions,
so
they
can
move
forward.
So
they
kind
of
move
out
it
like
they're
kind
of
crawling
along
and
they're
saying.
Is
that
there's
this
competing
process
where
they're
kind
of
retracting
and
these
filopodia
out
in
different
directions?
And
so,
if
you
have
this
happening,
sort
of
you
know
symmetrically.
B
B
B
So,
let's
see
what
they
say
about
it,
how
symmetry
breaking
manifests
itself
for
such
cells
is
therefore
elusive.
We
take
inspiration
from
the
scallop
theorem
proposed
by
purcell
for
micro
swimmers
and
newtonian
fluids,
so
this
is
just
like
a
regular
newtonian
fluid,
as
opposed
to
like
a
non-newtonian
fluid.
So
if
you're
familiar
with
non-newtonian
fluids
like
there's,
this
stuff
called
oobleck,
which
is
a
cornstarch
that
is,
you
know,
doesn't
have
the
sort
of
properties
of
of
water
say,
which
is
a
newtonian
floor
and
so
they're
using
a
simple
sort
of
medium.
B
That
is
where
these
swimmers
are
used
to
self-propelled
objects
undergoing
persistent
motion
that
low
reynolds
number
and
the
reynolds
number
is
a
num
coefficient
of
drag
more
or
less.
It's
like
when
the
cell
is,
if
there's
a
lot
of
resistance
or
little
resistance
to
the
flow
of
water,
the
reynolds
number.
B
So
if
the
object
creates
a
wake,
if
there's
a
lot
of
turbulence,
then
the
reynolds
number,
I
think,
is
high
if
there's
a
hard
flow
or
if
there's
a
turbulent
flow
and
if
it's
a
laminar
flow
or
if
there's
very
little
resistance,
it's
a
low
reynolds
number.
So
this
is
where
it's
just
kind
of
moving
through
a
laminar
flow.
B
We
report
similar
observations
for
cells
crawling
in
3d.
We
quantify
saw
motion
using
a
combination
of
3d,
live
cell,
imaging
visualization
of
matrix
displacement
and
a
minimal
model
with
multi-polar
expansion.
So
minimal
model
is
like
this
cell
is
just
modeled
very
minimally.
We
don't
model
every
process
and
we
just
have
a
very
simple
model
of
the
cell
and
it's
filopodi
and
it's
actin
filaments
within
there,
so
that
it
just
has
a
a
model
where
it's
generating
these
processes
and
moving
and-
and
things
like
that.
So,
okay.
B
B
D
D
B
B
D
B
Yeah,
well,
why
don't
we
go
through
the
rest
of
the
abstract?
Maybe
the
figures
and
see
what
they
have
here,
so
the
existence?
Okay,
I
think.
Actually
here
we
show
that
our
cells
embedded
in
the
3d
matrix
for
meiosis
and
driven
forced
dipoles
at
both
sides
of
the
nucleus,
so
that
you
know
that's
kind
of
like
where
you
have
these
dipoles,
so
they
kind
of
come
like
this,
these
meiosin
filaments.
B
B
So
it's
the
matrix
that
the
cell's
embedded
in
it's
pinching
outsides,
it's
pinching
forward,
or
you
know
it's
making
some
interact,
interface
of
the
environment
or
with
the
the
matrix
that
it's
in
the
existence
of
a
phase
shift
between
the
two
dipoles
is
required
for
a
directed
cell
motion
which
manifests
itself
as
cycles
of
the
finite
area
in
the
dipole
quad
pool
quadruple
diagram,
a
formal
equivalence
to
the
per
cell
cycle
and
talk
about
that
in
the
paper.
We
confirm
this
mechanism
by
triggering
local
dipolar
contractions
with
a
laser.
B
This
leads
to
directed
motion,
so
they
actually
trigger
these
dipole
dipole
contractions
they're
able
to
see
directed
motion
from
that.
Our
studies
revealed
that
these
cells
control
their
motility
by
synchronizing
dipolar
forces,
distributed
at
the
front
and
the
back
so
they're
able
to
control
the
synchronization
of
these
in
the
asynchronous
so
synchronization
and
time
where
they're
activated
this
result
opens
new
strategies
for
external
control
cell
motion,
as
well
as
the
design
of
micro
crawlers.
So
there's
a
lot
of
applications
to
this.
B
So,
let's
see
if
they
talk
about,
they
talk
about
two-dimensional
surfaces
and
then
three-dimensional
surfaces
and,
let's
see
if
they
talk
about
the
okay
here-
are
the
results.
So
here
are
the
pictures
here
of
this
contraction
and
extension,
where
there
are
contraction,
forces
and
extension
forces
that
are
countervailing.
This
is
the
matrix
displacement
chart.
So
this
is
the
cell
underneath
these
green
arrows,
these
green
arrows
are
the
the
forces
coming,
I
think,
from
the
cell
in
the
in
the
matrix.
B
D
B
B
B
Yeah,
hello,
yeah,
so
yeah.
This
is
the
setup
here
where
they
have,
they
do
a
lot
of
fluorescence
imaging
of
the
cell.
Of
course,
they
want
to
like
characterize
a
cell
they're
using
different
markers,
to
show
some
of
these
things
in
the
cell,
then
they're
visualizing,
these
flow
fields.
So
you
have
these
things
that
in
the
matrix
that
you
can
see
as
a
consequence
of
what
this
cell
is
doing,
the
kind
of
forces
it's
generating-
and
this
is
like
I
said
it's
a
slow
process.
B
This
isn't
this
these
visualizations
of,
say
like
meiosin
clusters,
which
are
muscle
proteins
that
aggregate
those
are
over
like
a
60
second
interval,
then
their
matrix
displacement.
You
can
see
over
a
90
second
interval
here
for
contraction,
but
these
images
here,
where
you
look
at
the
whole
cell,
you
know
you
can
image
the
syrup
to
25
minutes
and
it
gives
you
these
fields.
B
B
This
is
a
video
which
we
can't
play
because
it's
embedded
in
the
paper,
but
I
guess,
if
you're
on
the
web,
you
can
see
it
microtubule
asymmetric
distribution
is
associated
with
cell
polarity
during
motion,
so
they're
going
to
be
showing
about
the
distribution
of
microtubules
in
the
cell,
so
the
microtubules
on
our
cell
here
you
know,
are
embedded
in
the
in
the
in
the
membrane
here
and
in
in
around
there.
So
you're
going
to
have
a
lot
of
that
sort
of
thing
that
you
can
look
at
as
well.
B
Yeah
so
so
the
cell
deforms
a
derived
matrix
that
gives
you
kind
of
a
measure
of
where
it's
moving
and
how
it's
moving
and
then,
let's
see,
if
there's
anything
else,
that
we
want
to
look
at
in
the
paper.
So
this
is
the
sort
of
the
figure
here
where
they
show
this
contraction
extension,
and
so
it's
kind
of
moving
back
and
forth
and
you
can
actually
get
directed
movements
if
you
trigger
it,
it'll
go,
you
know,
it'll,
polarize
and
then
it'll
migrate,
so
the
cell
polarizes
in
this
direction.
Then
it
migrates
in
this
direction.
B
Yeah,
so
the
cell's
going
to
be
you
know,
the
cells
are
always
moving
like
they're
crawling
or
they're
migrating
to
one
place
to
another.
If
you
have
like
a
two-dimensional,
if
you
have
a
two-dimensional
cover
slide,
where
you
have
a
cell,
you
can
like
actually
look
at
it
like
migrating
across
the
surface.
If
you
have
like
a
a
chemical
gradient
or
something
it'll
move
in
the
direction
of
the
chemical
gradient,
you'll
see
it
like
kind
of
extending
these
little
processes
out
and
moving
towards
some
part
of
the
gradient.
B
So
they
do
this
all
the
time
and
they
also
do
it
in
in
in
the
tissue,
so
they're
always
migrating
around
in
development.
They
do
it
in
tissues
and
they,
but
they
have
this
directionality.
What
this
paper
showing
is
sort
of
the
origins
of
that.
So
you
have
a
like
an
origin
where
you
know
it
has
to
you
know
it's
starting
off
from
like
a
a
newly
divided
celse,
and
you
know
it
doesn't
have
to
be
spherical.
It's
usually,
you
know
some
sort
of
blob.
B
You
know
the
the
shape
is
sort
of
we're,
assuming
the
shape
is
they're
assuming
the
shape
is
spherical
here,
but
it
could
be
like
something
like
a
elongated
cell
or
something
and
some
cells.
They
have
like
this
weird
irregular
phenotype,
where
it's
kind
of
like
a
tree,
almost
like
fibroblasts
in
skin,
are
kind
of
long
and
thin,
and
they
have
this.
You
know
these
branching
processes
neurons
as
well,
so
these
cells
start
out
and
they
start
out.
You
know
with
these
processes.
B
They
also
send
out
these
processes
called
filopodia
which
extend
out
into
the
environment
and
help
them
move
they're,
almost
like
little
arms
that
come
out
and
pull
them
along
or
push
them
along,
and
so
the
thing
is
is
that
when
you
have
these
two
different
ends
of
the
cell,
like
two
different
poles
of
the
cell,
they
often
compete,
so
they
move
back
and
forth.
They
don't
really
have
a
directionality.
B
B
So
if
you
wanted
to
build
like
a
little
crowing
cell,
you
would
have
to
figure
out
to
take
that
cell
and
to
say,
okay,
you
know
it's
not
enough,
just
to
have
a
cell
and
maybe
like
put
out
some.
You
know
nice
chemical
gradient
for
it.
You
know
you
might
have
to
figure
out
how
to
get
it
to
move
in
that
direction
or
to
start
start
it
off
in
that
direction.
B
B
B
Physical
movement,
so
sometimes
if
cells
you
know,
are
on
a
certain
substrate
and
it's
a
certain
surface.
If
you
have
a
stem
cell
and
you
put
it
on
that
surface
and
it
starts
moving
around,
it
gets
feedback
from
those
forces
and
it
can
actually
force
certain
fates
to
or
can
force
certain.
You
know,
conditions
where
it
differentiates
in
a
certain
way.
B
There
have
been
a
number
of
studies
where
they've
shown
that
they
put
it
on
like
a
certain
surface
and
the
cell
is
extending
its
full
podium
feeling
the
surface
and
then
differentiating
like
accordingly.
So
that's
that's!
That's
stuff!
That's
like
you
know.
If
you
want
to
control,
if
you
want
to
take
stem
cells
and
control
the
fate,
you
can
do
something
like
that.
But
that's
that's.
D
D
D
B
B
This
is
by
ryan,
lavin
and
authors,
and
this
is
a
paper
from
frontiers
in
cell
and
developmental
biology
called
ivalve,
and
this
is
a
modular
python
based
model
for
simulating
developmental
eye
type
diversification,
so
eye
type
diversification
is
actually
quite
prevalent
in
evolution.
You
know
we
started
out
with
eyes
in
in
the
cambrian.
I
guess
where
you
had
organisms
that
started
moving.
So
when
organisms
started
moving
a
lot,
they
had
a
high
degree
of
motility
and
they
had
to
avoid
predators.
B
So
you
know
in
the
cambrian
explosion,
you
started
to
get
food
webs
where
you
had
predators
and
prey
and
they
were
moving
around
and
trying
to
avoid
one
another
or
catch
their
dinner,
so
they
started
to
get
eyes
and
that
that's
the
hypothesis
of
how
eyes
in
animals
started
to
evolve
and
then,
since
then
you
have
you
know
you
had
these
simple
light
sensors
at
that
point
and
then
as
they
evolved,
you
had
a
ver.
You
have
a
very
large
diversification
of
eye
types.
B
So
this
again
dick
sent
me
this
paper
and
so
vision
is
among
the
oldest
and
arguably
most
important
sensory
modalities
for
animals
to
interact
with
their
external
environment.
So
we
we
can
see
that
mounting.
Evidence
indicates
that
genetic
networks
required
for
visual
system
formation
and
function
are
relatively
well
conserved
between
species.
B
So,
aside
from
all
this
diversity
that
you
see
in
nature
amongst
animals,
in
insects
and
in
mammals
and
in
in
fishes
and
other
places,
there's
these
genetic
networks
that
control
some
of
these
visual
circuits
or
the
formation
of
eyes
is
well
conserved
between
species.
So
what's
happening
is:
is
that
they're
gene
regulatory
networks
that
are
sort
of
producing
a
lot
of
variants?
B
B
This
raises
the
question
as
to
how
common
developmental
programs
are
modified
functionally
in
different
eye
types.
So
they
approach
this
issue
through
ivalve.
This
is
an
open
source.
Python
based
model
that
recapitulates
eye
development
based
on
developmental
principles
originally
identified
in
drosophila.
B
B
Yeah
omittidia,
so
we
actually
did
a
paper
on
this
and
I'll
show
you
that
in
a
minute,
but
this
is
the
the
model
for
the
eye
in
in
actually
in
a
lot
of
insects,
but
drastic
in
particular.
Proof
of
principle.
Experiments
show
that
this
program's
animated
timeline,
successfully
simulates
early
eye
tissue
expansion,
neurogenesis
and
pigment
cell
formation,
sequentially
transitioning,
from
a
disorganized
pool
of
progenitor
cells,
which
are
like
stem
cells
or
early
neurons.
They're.
B
Not
really,
you
know
fully
formed
yet
they're,
just
kind
of
like
you
know
getting
started
as
the
celtics
are
going
to
be
to
a
highly
organized
lattice
of
photoreceptor
clusters,
which
are
these
cells
that
can
detect
light
or
cell
groups
of
cells
that
can
detect
weight
wrapped
with
support
cells.
So
you
have
these
cells
that
can
detect
light
at
different
frequencies,
and
then
you
have
support
cells.
B
So
if
you
look
at
a
retina,
for
example,
we
have
cells
that
have
you
know,
there's
like
a
relay
where
you
have
a
photo
sensitive
cell
and
there's
a
relay
back
to
there
or
from
it's
back
to
front
in
our
retina
to
a
circuit.
That's
going
to
go
to
the
central
nervous
system,
and
then
you
have
support
cells
in
interspersed
there,
which
are
important,
because
these
cells
are
using
a
lot
of
glucose
they're,
burning,
a
lot
of
energy
and
there's
a
lot
of
maintenance
that
needs
to
take
place.
B
So
these
support
cells
are
useful
further
tweaking
just
five
parameters:
precursor
pool
size,
which
is
the
number
of
precursor
cells.
You
start
with
founder
cell
distance
and
placement
from
edge,
which
is
the
first
cell
that
shows
up
in
this
omittidia
and
placement
from
the
edge
of
the
area
that
you
want
to
simulate,
photoreceptor,
subtype,
number
and
cell
death
decisions.
B
So
this
is
where
you
have
apoptosis,
where
sometimes
cells
die,
they're
programmed
to
die
at
a
certain
point
in
development,
or
sometimes
they
get
they
get.
You
know
they
die
because
they
get
injured
or
they
get
damaged
or
something,
but
they
undergo
this
program
cell
death.
This
is
important
to
include
in
the
model
as
well.
B
Thus
ivolve
sheds
light
and
common
principles
of
eye
development
and
provides
a
new
computational
system
for
generating
specific,
testable
predictions
about
how
development
gives
rise
to
diverse
visual
systems
with
a
commonly
specified,
neural
epithelial
ground
plan,
so
neural
epithelial
ground
plan
just
means
you
know
what
are
the
from
that
genetic
circuit.
What
are
the
cells
that
are
specified
for
that
sort
of
architecture?
B
D
D
D
B
A
A
B
So
I've
been
trying
to
zoom
in
on
this,
but
it's
hard
well,
I
can
show
what
I
should
do
here,
but
this
is
the
interface
that
they
have
for
their
program.
This
is
in
a
mac
eye
development
model,
so
they
it's
like
an
agent-based
model.
Basically,
so,
if
you're
familiar
with
that,
it's
very
similar,
you
have
these
agents,
which
are
the
cells,
and
you
have
this-
these
parameter
values
over
here.
So
you
can
change
them.
B
You
have
you
know
the
maximum
number
of
cells
to
start
with
the
growth
rate,
the
furrow
velocity
furrow.
Is
this
thing
here
that
moves
across
this
array?
I
can
show
you
pictures
of
it
in
our
in
our
diagrams
that
we
did
for
a
paper
that
we
did
on
omatidia
expansion
and
development,
but
this
is
basically
the
simulation.
B
B
It
moves
in
one
direction
or
another
and
it
just
kind
of
transforms
these
cells.
You
have
these
different
factors
here
and
then
you
have
apoptosis
of
unused
cells,
which
is
this
programmed
cell
death
that
removes
those
cells,
and
you
end
up
with
this
pattern
formation
after
the
cells
that
aren't
useful
die
off.
So
then
you
have
all
these
different
things
here,
where
you
have
these
different
selection
processes,
exclusion,
radiuses,
so
they're
doing
all
this
stuff.
B
D
D
B
Yeah
they
say
that
they're,
they
say
tweaking
just
five
parameters,
so
I
guess
they
probably
tried
a
larger
number,
but
they
know
that
these
five
parameters
actually
do
the
best
job
of
predicting
a
lot
of
this
diversity.
So
you
know
there
are
other
parameters
that
might
be
somewhat
useful,
but
not
as
predictive
as
those.
C
B
I
am
so
this
is
a
little
bit
better
view
of
the
morphogenetic
furrow,
and
this
is
the
this
is
the
formation
of
the
eye.
This
is
so
this
is
your
depian
staining
of
the
developing
adult
eye
in
the
late
third
larval
instar.
So
this
is
a
larval
stage
of
drosophila
where
this
eye
is
being
formed.
This
is
an
imaginal
disc,
and
these
are
the
omittidia,
and
this
is
the
morphogenetic
furrow,
this
red
mark
here
and
this
thing
moves
across
the
surface.
B
You
know
it
has
this
this
sort
of
u-shaped,
and
if
you
you
know,
I
don't
know
if
you
can
see
it
from
here
directly,
but
you
can
see
the
line
and
it
moves
across
then
there's
this
differentiation
process
that
goes
on
v
is
this:
is
a
scanning
electron
microscope?
So
this
is
much
higher
resolution
than
the
dappy
image.
Dappy
is
the
stain
that
they
use
to
visualize.
Some
of
this.
The
electron
microscope
shows
some
of
this
surface,
illustrating
the
precise
and
regular
positioning
of
these
omittidia.
B
B
It's
this
simulation
of
b
is
regular
spacing,
so
you
can
see
that
there's
a
differentiation
process
going
on
here
and
then
this
is
an
adult
like
a
a
mature
version
of
the
eye
with
the
ohmatitia
aligned,
and
this
is
what
they
look
like
in
the
simulation,
so
they
develop
progressively
from
right
to
left.
This
is,
I
think
this
is
backwards
with
respect
to
this,
but
that's
that's
basically
what
it
is
I'll
go
ahead.
D
D
B
B
B
Versions
of
of
arthropod
eyes,
so
these
are
authentic
eye
layouts.
This
is
a
is
a
male
twisted
wing.
This
is
a
type
of
mutant
in
drosophila
called
twisted
wing.
The
mutants
have
different
names
like
that,
but
this
is
one
of
the
mutant,
phenotypes
and
n,
a
which
have
chunk
vision,
which
have
relatively
large
units
with
a
small
image
forming
eye
b
is
a
la
lepidoptera,
which
is
a
butterfly
and
eyes
that
they
have.
They
have
their
larvae
tend
to
have
six
material
like
units
that
are
spread
far
apart.
B
D
B
D
B
So
the
single
lens
eye
that
likely
evolved
from
an
ancestral
compound
eye,
so
our
eyes
actually
may
have
evolved
from
an
ancestral
compound
eye,
and
you
can
see
this
in
sophie
larvae,
which
is
a
different
species
here,
where
they
have
a
single
eye,
and
you
can
see
some
of
the
things
that
they
can
simulate
here,
where
they're
simulating
this
mode
of
development
they're
playing
around
with
how
these
things
develop
and
they
get
this
sort
of
output
here.
And
so
it's
very
you
know.
B
D
B
B
B
B
Getup
repo
here
I
developmental
model,
bushbeck
lab,
and
this
is
yeah.
This
is
it
here.
Thank
you
for
that
elon.
So
this
is
an
example
here
of
the
eye
development
model
and
you
can
set
it
up
in
python,
install
pycharm
and
there's
some
other
things.
You
need
to
do
to
get
it
working,
but
it
should
work
so
that
that's
that's
eyeball.
B
This
is
our
paper
that
we
did
morphogenetic
processes
as
data.
So
we
what
we
did
was
we
had
this
image
of
a
drosophila
imaginal
disc,
which
is
this
disc,
that
I
showed
you
in
one
of
the
figures
where
all
these
omittidia
were
developing
on
and
so
drosophila
has
a
number
of
imaginal
discs
that
it
uses
in
development
to
develop
wings
and
eyes.
B
B
D
B
Okay,
all
right
so
yeah,
so
this
the
the
point
was,
is
that
this
was
drawn
hand
drawn
from
looking
under
a
microscope
at
this.
So
you
have
this
hand-drawn
image
and
the
benefit
of
a
hand-drawn
image.
Well,
it
may
not
be
as
precise
as
like
a
microscopy
image.
It
does
give
you
some
additional
resolution,
so
we
could
blow
it
up
to
a
larger
size
and
actually
then
digitize
the
drawing
and
then
it
would.
B
You
know
you
could
pick
out
some
of
these
patterns
at
a
pretty
high
resolution,
so
this
is
actually
the
there
was
a
drawing,
and
then
these
these
figures
are
some
of
the
segmentation
that
was
done
on
it.
So
we
did
a
lot
of
segmentation
in
in.
I
think
it
was
fiji,
which
is
an
open
source
pro
program
for
analyzing
microscopy
images.
They
have
a
lot
of
tool
kits
and
their
tool
boxes
for
segmenting
images.
This
was
a
segmentation
of
this
omittidian.
B
We
had
something
like
I
think,
8
000
cells,
that
we
were
able
to
segment
out
from
here,
so
we
were
able
to
segment
eight
thousand
and
identify
their
centroids
and
their
positions
within
a
coordinate
system.
So
then,
you
can
actually
ask
the
question:
what
kind
of
patterns
do
these
things
form
the
the
image
itself
had
the
furrow,
and
you
can
see
in
this
image.
The
furrow
was
here.
You
had
some
undifferentiated
cells
here
and
then
the
differentiated
cells
over
here,
so
this
furrow
was
moving
across
the
image
from
left
to
right.
B
This
is
the
differentiated
part.
This
is
the
undifferentiated
part,
so
you're
actually
catching
it
in
the
active
differentiation
with
the
furrow,
and
you
could
say
when
the
fir
when
things
are
differentiated,
what
are
the
patterns
of
these
umitidia
and
then,
when
they're
undifferentiated?
What
are
the
patterns?
And
you
can
do
a
lot
of
things
with
these
data?
It's
very
good.
So
this
is
yeah,
although
exactly
yeah
yeah.
D
B
D
B
So
this
is
this:
was
there
were
a
lot
of
work
on
this
like
artistically
before
we
got
to
do
the
analysis
on
it?
So
this
shows
you
how
you
know
what
kinds
of
data
you
can
use
and
then
this
is
a
close-up
of
the
image.
So
this
is
just
the
raw
image.
You
know
the
this.
These
are
the
where
you
had
everything
closed
up
and
you
had.
You
know
this
is
digitized
as
an
image
and
it's
ready
to
segment.