►
From YouTube: DevoWorm (2021, Meeting 21): GSoC Activities Update #1, Geometry of Cells, Embryos from Input Cells
Description
Update on Summer of Code activities (Mainak Deb, Improving DevoLearn), OpenWorm poster for the International C. elegans Conference, short lecture on Cell Shape Geometries in Archaea. Papers on synthetic development and embryogenesis via gastruloid. Attendees: Jesse Parent, Richard Gordon, Bradly Alicea, Shruti Raj Vansh Singh, and Mainak Deb.
B
A
Yeah
so
yeah,
I
know
I
know
there
was
a
problem
with
the
timing.
I
had
to
change
the
time
quickly
before.
A
D
A
Okay,
there
we
go
so
today,
I'm
gonna
talk
about
a
couple
things.
I
wanted
to
go
over
a
couple
things
while
we're-
and
I
know
no
one's
really
here,
but
first
of
all,
congrats
to
shruti
and
krishna
for
their
acceptance
in
the
neuro
match
academy.
So
they've
been.
A
So
they've
been
accepted
in
the
narrow
match
academy
for
the
summer
surety's
been
accepted
as
a
attendant
and
christian
has
been
accepted
as
a
ta.
So
that's
he's
gonna
be
the
ta
for
the
deep
learning
group
and
then
surely
I
guess
is
going
to
be
attending
the
core
mma
session.
So
there's
an
anime
core
session
which
is
july.
2021
then
there's.
I
don't
know
I've
not
seen
this
before
this
minis
thing,
which
is
a
half
a
full
day
workshop.
A
So
I
don't
know
what
that
is
and
then
there's
the
deep
learning
session,
which
is
deep
learning
curriculum
for
three
weeks.
This
is
what
krishna
is
going
to
be
a
ta
for
so
congratulations
to
krishna
and
surety.
A
A
The
movement
validation
group,
so
what
they're
doing
is
they're
taking
videos
of
different
experiments
and
they're,
putting
them
into
this
platform
they're,
taking
the
data
they're,
putting
it
into
the
platform
and
they're
sort
of
aggregating
it
for
people
to
use.
So
there's
a
lot
of
interesting
data
in
here
there's
there
the
samples
are
hosted
on
youtube.
You
download
a
video.
You
can
then
go
ahead
and
analyze
it.
A
They
have
a
lot
of
summary
data
here,
but
they
have
you
can
I
mean
I
think
I
don't
know
if
my
oak
yeah
man,
my
oak,
may
have
used
it
last
year
for
some
of
his
stuff.
It
may
have
been.
You
know
it
might
be
useful
to
my
knock.
It
might
be
useful
to
other
people
working
on
some
of
the
stuff
for
the
worm,
just
in
terms
of
machine
learning
alone,
it's
kind
of
a
valuable
resource.
A
I
know
there
was
a
paper
published
on
this
recently
where
they
were
using
some
of
this
data
and
it
wasn't
an
open
worm,
affiliated
group,
but
they
used
a
lot
of
the
data
from
this
database
to
do
their
paper,
and
I
don't
have
the
paper
on
open
right
now,
but
I
can.
I
can
pull
it
up
for
next
week
and
we
can
talk
about
it.
So
that's
something
else
to
keep
in
mind.
This
is
movement.openworm.org
just
in
case.
You
want
to
know
more
about
it.
A
You
all
right,
so
that's
that's
for
that.
I
also
wanted
to
mention
that
we
have
our
poster
that's
being
submitted
to
the
international
c
elegans
meeting,
and
this
is
the
open
arm
poster.
So
this
is
a
poster
that
has
all
the
different
sub
projects
in
open
worm,
not
all
of
them,
but
like
a
sampling
of
them.
We
have,
for
example,
the
cybernetic
project
which
we
talked
about
last
week.
We
have
the
introduction,
which
is
just
an
introduction
to
the
platform
and
the
project
in
the
foundation
and
the
different.
A
You
know
just
the
basic
information
talking
about
the
docker
file.
Then
you
have
this
data
management,
which
is
a
initiative
which
is
openworm,
meta
and
channelworm,
so
they
do
a
lot
of
like
data
management,
formatting
data,
and
then
there
are
the
neural
simulations
that
are
neuroml
is
a
platform
that
they
use
for
this.
So
c302
is
the
the
semi
neuronal
simulation
platform
and
neuroml
is
used
as
sort
of
a
basis
for
a
lot
of
that
work.
A
In
a
little
bit
more
so
a
lot
of
this
stuff
is,
you
know
this
is
how
the
the
current
docker
file
looks.
So
you
have
this
docker
image
which
you
can
download
to
your
computer.
You
can
open
it
up
and
run
it,
and
these
are
the
current
things
that
it's
running
so
it's
running
on.
You
have
to
run
it
on
ubuntu
or
some
linux
platform.
A
You
need
opencl
and
python,
and
then
you
run
the
following
simulations.
You
can
run
a
cybernetic
simulation,
you
can
run
c302,
you
can
run
ow,
meta,
pi,
neuro
ml
and
neuron,
and
so
you
get
these
two
are
languages
that's
dependent
on,
and
these
three
things
are
what
what's
running
and
so
you're
going
to
generate
for
cybernetic
and
c302.
I
think
you're
generating
a
bunch
of
images
that
end
up
getting.
There
have
been
some
problems
with
getting
the
docker
container
to
run
without
like
being
obtrusive
in
terms
of
memory.
A
So
but
these
these
two
simulations
generate
a
bunch
of
images
that
are
saved
on
the
local
machine
and
you
can
actually
go
through
these
simulations
just
by
pressing.
You
know
pressing
a
button
so
that
that's
a
nice
feature
of
our
docker
container
and
again,
you
know
I'd
like
to
put
a
diva
worm
component
in
here,
but
I
re,
I
don't
think,
we've
really
found
the
right
component.
We've
talked
about
simulating
like
having
a
simulation
of
the
embryo.
A
We've
talked
about
having
other
types
of
things
that
are.
You
know
that
we
might
end
up
using,
but
that
that's
going
to
be.
A
You
know
that
that's
a
we
can
continue
that
conversation
later
so
then,
there's
gepetto,
which
is
the
platform
that
enables
visualization
and
simulation
neuroscience
models.
So
this
is
a
bit
broader
than
the
neuroscience
models
that
we
use
in
c302.
This
is
actually
something
that
we
can
integrate
with
things
like
drosophila
data
and
other
types
of
things,
so
geppetto
is
actually
run
by
yeah.
It's
kind
of
semi-independent
of
open
worm
as
well,
and
the
worm
sim
website
is
where
you
want
to
go
to
find
out
more
about
that
then.
Finally,
our
contribution
is
here.
A
This
is
the
diva
worm
contribution.
This
is
where
I
just
kind
of
talk
about
the
different.
I
think
they've
put
xxx
in
here.
They're
going
to
be
renumbering
the
references,
but
so
you
know
in
this
one
I
put
a
little
triangle
of
stuff,
so
this
is
the
connection
to
open
orum.
This
is
the
tivor
group,
and
this
is
the
open
data
component.
Then
I
put
this
component
in,
which
is
the
simulations
that
we
have
from
diva
learn.
A
We
have
a
characterization
of
our
five
tuple,
our
five
dimensional
data
space,
the
3d
space
plus
time
and
context
this
theta,
which
can
be
a
number
of
things,
and
I've
talked
about
that
in
previous
meetings
and
then
the
divo
learn
logo.
So
we're
you
know
talking
about
divo,
learn
we're
talking
about
some
of
the
work
that
we've
done
with
comparative
biology
and
then
complex
networks,
which
is
there's
an
image
in
here.
It's
hard
to
see
at
this
magnification,
but
there's
a
embryo
network
in
there.
A
So
and
then
I
put
a
couple
of
references
in
here
so
hopefully
that's
a
good
representation
of
diva
worm
to
the
open,
or
this
is
to
the
international
c
elegans
meeting,
and
this
is
going
to
take
place
online.
I
think
this
month,
maybe
two
more
weeks
until
it
happens.
So
hopefully
they
res
it's
well
received.
A
I
did
a
diva
worm
poster
at
that
conference.
I
think
in
2015
and
actually
I
met
up
with
steve
larsen
there.
It
was
held
in
los
angeles
and
he
was
in
san
diego
at
the
time,
so
he
just
drove
up
and
then
he
we
met
there
and
that's
the
first
time
I
had
met
him
in
person
and
then
I
met
him
again
in
the
same
year
at
the
at
another
conference,
so
that
was
I
haven't
seen
him
since
then,
so
I
in
the
flesh.
A
So
that's
I
don't
know
we
still
won't
be
able
to
meet
in
person.
So
that's
the
thing
there,
okay,
so
so
that's
all
I
have
for
that
now.
I
wanted
to
invite
my
knock
to
give
a
update
on
his
his
stuff.
E
Yeah
I'll
just
share
my
screen
now:
okay,
I'll!
Do
it
right
now?
Okay!
So
so,
can
you
see
my
screen?
Yes,
hello,
okay,
so
first
I'll
go
over
the
weekly
update
and
then
I'll
go
over
the
coding
period
angling
that
we
have
okay,
so
okay
I'll
get
into
it
now:
okay,
so,
okay!
So
this
week
I
basically
spent
a
significant
number,
a
significant
amount
of
time
trying
to
find
some
new,
some
new
data
to
work
with
pc
and
I
actually
browse
through
a
lot
of
papers.
E
E
E
E
It
for
future
use
so
I'll
go
ahead
and
get
into
the
details
of
the
of
the
data,
so
the
data
basically
contains
feature
and
label
pairs
like
this,
and
this
image,
basically,
is
the
fluorescence
microscopy
data
and
to
the
right.
We
can
see
the
segmentation
map
and
it's
actually
semantically
segmented,
which
is,
of
course,
which
is
a
good
thing
and
the.
E
E
Classify
these
empty
points,
so,
okay,
so
as
I
had
said
that
the
data
was
in
in
a
tiff
format,
so
I
basically
have
to
convert
them
into
a
jpeg
format
for
ease
of
use
and
I
stored
them
in
this.
This
drive
website
this
drive
link
and
there's
also
the
collab
notebook,
and
so
if
anybody
wants
to
check
out
the
code
and.
E
So
what
I
plan
with
this
data
is
is
to
basically
train
a
neural
network
to
take
input
of
3d
data
in
form
of
these
slices,
and
then
I
have
have
the
model
basically
map
it
to
these
3d
segmentation
maps.
I
know
it's
a
bit
far
fetched
now,
but
I'll
see
I'll,
see
what
I
can
do
so
coding
period,
actually,
okay,
so
okay,
so
I
guess
that's
it
for
this
week's
update.
E
E
Yeah
yeah,
okay,
so
the
first
week
which
starts
today
so
so
for
the
first
week,
I
basically
have
plans
to
refactor
all
the
models
according
to
the
inference
engine
class
and
then
inspect
the
data
sets
then
take
some
feedback
if,
if
like,
if
someone
has
to
add
anything
I'll,
just
go
ahead
and
get
in
touch
and
also.
E
E
E
E
C
E
Which
would
be
which
would
enable
the
users
to
run
inference
on
numpy
eyes?
And
I,
like
I,
found.
E
E
In
the
first
week,
which
is
week,
seven
I'll
I'll
explore
the
epic
two
epic
two
data
set,
but
I
think
after
I
found
the
cell
tracking
challenge
data.
I
think
that
could
also
be
replaced
by
this,
like
I
could
also
work
on
the
like.
I
could.
E
With
that
data
set
so
I'll
see
what
what
is
to
be
done
there
and
in
the
following
weeks,
I'll
explore
some
some
dimensionality
reduction
techniques
like
a
umap
and
some
other
projection
techniques
and
I'll
try
to
extract
some
patterns
or
some
some
sort
of
a
meta
info.
That.
C
E
C
E
The
readme
file,
so
I
think
it
requires,
read
the
docs
for
like
I
think
it
requires
a
formal
documentation
on
the
internet
and,
I
think,
read
the
docs
or
get
book
could
be
used
for
that.
That
could
also
be
done
and
yeah
the
final
week
I
have.
I
have
left
a
couple
of
days
in.
C
E
A
Hello,
shorty
and
jesse.
How
are
you
that's
very
good?
I
want
to
talk
about
a
couple
of
the
slides,
so
if
you
could
go
to
number
two
yeah,
okay,
so
hunting.
C
A
Data
sets,
so
you
said
that
there
was
this
cell
tracking
challenge
yeah.
What
are
the
data
sets
there?
Can
you
show
and
you
go.
E
E
There
is
data,
because
okay,
so
these
are
the
data
sets
that's
available
like
I
am
not
really
familiar
with
all
these
other
data
sets.
But
what
I
found
interesting
is
this
one,
so
this
is.
E
E
A
E
E
E
E
E
What
we
have
here
is
that
for
each
time
point
we
have
these,
we
have
these
slices,
so
this
essentially
this.
This
essentially
is
for
one
single
time
and
these
35
slices.
C
E
A
This
further
and
I'll
keep
you
guys
updated
with
that
yeah.
It
would
be
great
yeah,
that's
great,
and
so
it's
from
the
same
group,
so
it
gives
they
use
sort
of
the
same.
C
C
A
Variation
between
groups
and
but
yeah,
it's
so
that's
good,
and
then
we
have
the
semantic
map,
which
is,
of
course,
valuable
and
then
the
data
here.
So
we
have
the
segmentation
maps
yeah.
Those
are
great.
You
talked
about
this
in
the
slack
channel,
so,
okay.
A
E
E
A
Great
then,
you
have
the
what
else
is
the
second.
C
E
A
Last
slide:
oh
your
schedule
right.
So
this
is
the
timeline
yeah.
So
I
guess
that
looks
good.
You
know
it's
going
to
be.
The
thing
you're
going
to
have
to
watch
out
for
is
you're
going
to
have
a
lot
of
things
in
one
week.
So
you
know
you
want
to
make
sure
that
you
kind
of
you
want
to
keep
track
of
time,
but
you
also
want
to
want
to
get
stuck
on
anything.
A
C
A
A
A
hub
for
users,
so
this
is
something
that
it
so
it
would
be
like
a
blog
out
of
github
I
o,
or
would
this
be
something
that
is
separate
because
we
have
a?
We
have
a
github
io
presence
already.
I
don't
know
we
might
want
to
keep
it
on
that
in
that
space
like
we
have
the
diva,
the
devo
zoo
and
some
of
the
other
stuff
there.
A
Yeah
yeah,
I
think
that
the
idea
of
having
these
little
demos
of
sliders
is
great.
We
didn't
we
haven't
gotten
to
that
stage
really
yet,
which
is
you
know
because
basically
yeah,
but
I
think
that
would
be
great
to
have,
especially
as
an
interactive
tool.
We
try
to
do
that,
sometimes
in
in
the
notebooks,
but
sometimes
those
are
not
they're
not
always
accessible
to
people,
so
I
think
yeah
doing
a
web
interface
would
be
good.
A
C
E
From
from
the
other
developers
in
this
community
because,
like
I'm,
not
very
experienced
in
bed
but
like
I'll
try,
my
best,
that's
all.
A
A
And
dick
wants
to
have
it
it
you
know,
wants
to
go
over
some
of
this,
maybe
yeah.
So
that
would
be
great.
Let's
see
so
now
I'm
gonna
is,
is
anyone
like
does
jessie
or
shrewdy
or
dick
want
to
present
in
anything
today
or
do
we
have
I'm
going
to
move
on
to
other.
B
A
Yeah
again,
congratulations
to
shruti
for
getting
into
aeromatch
yeah.
B
Yeah
I
I
could
give
a
very
brief
presentation
on
the
problem
of
segmenting
images
of
archaea
and
comparing
them
to
shape,
drops.
B
B
B
These
first
of
all,
let
me
explain
if
you
take
oil
in
water
and
shake
it
up.
Ordinarily,
you've
got
spherical
drops
okay,
but
if
you
have
a
little
bit
of
surfactant
present
and
you
cool
it
swollen
you
get.
These
polygonal
shapes.
That's
in
the
first
column
here,
they're
called
shape
droplets.
You
can
see
that
can
be
triangles.
B
They
can
be
triangles
with
concave
edges.
They
can
sometimes
form
these
weird
things
on
the
side.
They
can
be
squares.
Rectangles,
poly,
pentagons,
hexagons
and
they've
been
observed
up
to
nine
sided.
B
One
two,
three
four
and
you've
got
pentagons
in
higher
order.
Here's
a
12.
B
B
For
example,
if
you
look
at
the
archaic
carefully,
you
can
see
that,
even
though
this
is
triangular
instead
of
a
60
degree
corner,
they
seem
to
have
90
degree
quarters.
So
there
might
be
some
differences
in
that
regard.
There
might
be
some
structure
in
quarters
that
makes
the
90
degrees
okay.
So
here's
here's
the
problem.
B
I've
collected
a
large
number
of
pictures
of
these
things,
and
can
we
quantitate
the
shapes
and
compare
them
and
see
if
the
shape?
How
different
are
the
shaped
crops
in
general
from
the
archaea
and
how
similar
okay?
So
it's
a
it's
a
it's
a
large,
it's
a
large
segmentation
and
geometry
problem,
and
you
can
see
that
the
quality
of
the
images
is
very
different.
B
A
That
would
be
great
okay,
that's
it
yeah
yeah.
So
it's
yeah.
That
does
remind
me
of
the
stuff
that
we're
trying
to
do
with
the
talk
on
the
toponets,
where
we
have
these
geometric
shapes
that
you
know
we're
considering
like
the
sort
of
the
form
and
the
edges
instead
of
the
nodes
as
like,
informative,
and
so
you
know
to
fit
that.
It's
almost
like
you
know,
you
could
use
an
approach
to
like.
A
Finite
element,
analysis
or
something
similar
where
you
have
all
these
different,
you
know
we're
trying
to
fit
a
mesh
or
something
to
it
to
see
what
the
differences
in
shape
are.
B
B
A
Yeah,
if
people
are
interested,
that
would
be
great,
it
would
be
something
I
think
we
could
pick
up
on
a
theme
that
would
be
really
productive.
So
yeah,
that's
great.
Let's
see
we
have
some
things
in
the
chat:
okay,
yeah.
So
that's
linked
to
inox
slides
this
week's
blog
post,
which
is
something
that
my
knock
has
to
do
weekly
and
then
can
you
share
this
paper?
Would
love
to
read
it
so
yeah
the
paper
on
the
that
you
were
just
showing
on
the
on
the
archaea.
B
A
So
yeah,
so
all
right,
then
let
me
get
on
with
the
oh
here
we
share
your
screen.
Okay,
we
can
get
on
with
the
submissions
and
things.
This
is
our
submissions
list
as
you've
seen
before
we
have.
I
think
we've
cleared
a
fair
amount
of
this,
but
we
still
have
other
things.
We
need
to
revisit
so
number
eight,
this
open
one
poster.
I
just
discussed
this.
This
is
to
be,
I
think,
presented
in
about
two
weeks,
so
I
think
we're
pretty
much
done
with
the
edits
on
that.
A
I
think
that
was
a
good
summary
of
divo
diva
worm
as
it
stands
now
and
we
just
didn't
you
know
we
didn't
have
a
lot
of
space
to
go
into
things
in
depth.
So
that's,
but
that's
just
you
know,
to
give
people
a
taste,
and
I
wanted
to
put
a
lot
of
images
on
there
so
that
people
can
you
know
in
in
posters.
You
need
to
use
a
lot
of
images,
this
stuff
for
networks,
2021,
that's
still
kind
of
in
progress.
A
That's
going
to
be
a
push
for
that
later
on
this
month,
and
then
I
I'm
sorry
about
this
non-oral
cognition
paper.
This
is
something
that's
been
really
kind
of
long
suffering
and
I'm
trying
to
finish
it
up,
but
it's
been,
you
know
we're
trying
to
generate
a
workable
paper,
and
so
I
talked
to,
I
think
dick
and
some
other
people
about
it
this
weekend
and
I
just
sent
dick
a
version
of
it
and
it's
pretty
it's
it's
somewhat
rough,
but
I
wanted
to
get
some
feedback
on
it.
A
So
that's
that's
going
to
be
worked
on
as
well
this
month.
Could
you
potentially
send
me
a
version
of
the
paper
too
yeah?
Why
don't
I
do.
E
A
I'll
send
you
a
link
to
it
after
the
meeting
yeah
I
mean
it's,
it's
pretty
rough,
but
yeah.
If
you
could
provide
feedback,
that
would
be
great.
We
could
we
could
follow
up.
A
I
mean
it
really
needs
a
lot
of
fleshing
out
there's
a
lot
of
stuff
in
there
a
lot
of
ideas,
some
of
them,
maybe
aren't
great
for
the
scope
of
the
paper.
Some
are
maybe
very
important
and
need
to
be
expanded
upon.
So
we
need
to
figure
out
what
those
are,
and
so
that's
that
and
then
this,
let's
see
what
else
is
coming
up.
A
A
We
have
these
here,
the
test
of
williamson
symbiosis
and
the
molecular
level
simulations
of
diatoms.
Those
are
still
on
the
list
if
people
are
interested
and
then
this
actually
28,
we
have
on
the
list
which
dick
just
talked
about
with
a
quantitative
quantitative
comparison
of
our
key
and
shape
droplets.
A
So
just
to
keep
on
top
of
that.
That's
good!
What
else
do
we
talk
about?
Why
don't
we
just
get
into
papers
now,
since
it
gives
us
a
little
bit
more
time
to
talk
about
this,
but
so
got
a
couple?
Several
nice
papers
here,
I'm
always
trying
to
catch
up
on
on
the
newer
papers,
but
this
is
a
good
opportunity.
A
So
the
first
one
I'm
going
to
talk
about
today
is
the
synthetic
living
machines,
a
new
window
on
life,
and
this
is
a
I
think,
it's
a
new
paper
for
michael
levin
and
collaborator
yeah.
This
is
a
newer
paper,
so
you
know
he
does
this
stuff
with
xenobots,
which
are
these
robots
that
are
cells
essentially
and
they're,
using
synthetic
biology
to
sort
of
program,
those
cells,
and
so
that's
it.
That's
a
big
area
that
has
gotten
a
lot
of
press.
A
So
I
think
this
builds
off
of
that
they're,
making
the
argument
here:
synthetic
living
machines,
a
new
window
on
life,
so
summaries
increase,
control,
biological
growth
and
form
as
an
essential
gateway
to
transformative
medical
advances,
repairing
of
birth
defects,
restoring
lost
or
damaged
organs,
normalizing
tumors
all
dependent
understanding
how
cells
cooperate
to
make
specific,
functional,
large
scale
structures,
and
we've
talked
about
that.
With
respect
to
development-
and
you
know
regeneration
despite
advances
in
molecular
genetics,
significant
gaps
remain
in
our
understanding
of
the
mesoscale
rules
of
morphogenesis.
A
So
what
mesoscale
means
is
that
it's
not
like
the
interactions
between
molecules
and
it's
not
the
type
of
like
tissue
formation,
but
it's
this
mesos
or
or
you
know,
morphogenesis
over
the
course
of
all
of
development.
But
these
sort
of
meso
scale
rules
which
are
in
between
those,
and
so
we
don't.
You
know
we.
We
don't
really
understand
a
little
bit
of
how
it
works,
but
not
entirely.
A
So
this
is,
you
know
this
is
just
creating
sort
of
new
model
systems
instead
of
relying
on
the
typical
model
systems
that
we
use.
Here.
We
review
recent
advances
in
the
emerging
field
of
synthetic
morphogenesis,
the
bio-engineer
engineering
of
novel
multicellular
living
bodies
emphasizing
emergent
self-organization
tissue
level,
guided
self-assembly
and
active
functionality.
This
work
is
the
essential
next
generation
of
synthetic
biology.
A
Aside
from
useful
living
machines
for
specific
functions,
the
rational
design
and
analysis
of
new,
coherent
anatomies
will
greatly
increase
our
understanding
of
foundational
questions,
so
they
get
into
this.
So
here's
here's
figure
one
and
they
kind
of
talk
about
again.
This
is
one
of
these
broad
diagrams
that
kind
of
explain
what
they're
getting
at
here,
so
they
they're
sort
of
founding
this
new
area
here
synthetic
functional
morphologies.
A
So
you
know
this
requires
a
number
of
different
disciplines
it
requires
biology,
requires
us
to
understand.
Biophysics,
developmental
biology
and
even
evolutionary
biology
requires
computer
science.
In
this
case,
machine
learning
and
swarm
robotics
and
swarm
robotics
is
you
know
where
you
take
a
bunch
of
robots
and
you
try
to
make.
You
know,
try
to
get
them
to
behave
like
social
insects.
Basically
they'll,
you
know
communicate
with
one
another.
They
have
these
means
of
like
forming
swarms
at
a
desired
time
around.
A
Maybe
you
know
you
provide
them
with
computational
resources,
you
provide
them
with
like
cues
in
the
environment
and
they
behave
like
a
collective
and
so
that
that's
important
machine
learning
is
also
important.
Probably
other
areas
that
aren't
mentioned
here
also
synthetic
morphology,
cybernetics
and
cognitive
science
and,
of
course
we
know
mike
levin's
view
of
cognitive
science
and
in
cybernetics
and
how
that
relates
to
biology.
We
talked
about
that
last
week.
A
A
A
A
You
can
use
these
kind
of
models
for
drug
discovery
and
disease,
modeling
decoding
design,
principles
of
human
development,
the
evolution
of
body
plans,
which
we've
talked
about
in
the
past:
regenerative
medicine
and
birth
defects,
injury
and
cancer,
new
ai
platforms
and
algorithms,
swarm,
robotics
and
useful
biobot
functions,
and
so
these
are
all
things
that
you
could
use.
These
are,
very
you
know,
speculative,
especially
the
swarm
robotics
aspect,
where
you're,
basically
using
it
to
get
cells
to
collaborate
with
one
another
and
communicate.
A
A
A
So
we
have
a
synthetic
embryos
or
these
embryo-like
entities,
they
call
them
blastoids
and
they're
co-stained
with
f
act
in
here.
You
can
see
in
nano
green,
which
is
in
here.
So
you
can
see
in
this.
I
think
in
this
figure
here,
nanog
and
f
actin
are
expressed
in
different
parts
of
the
synthetic
embryo.
A
A
Then
they
show
this
here,
where
it's
a
schematic
to
show
strategy
for
generation
of
mouse
embryonic,
like
structures
combining
the
mouse,
which
is
the
red
component
here,
and
the
tscs,
which
are
the
blue,
which
are
these
specific
type
of
stem
cell
here.
So
these
are
mouse
esc's.
These
are
tscs,
and
these
are
the
so
you're
culturing
these
different
types
of
cells
in
different
contexts
and
then
they're
combining
them
into
a
single
organism
or
a
single
entity.
A
I
should
say-
and
then
they
have
these
gastruloids
at
120
hours
at
the
bottom,
after
aggregation,
showing
features
of
a
mouse
embryo
on
embryonic
day,
one
and
then
you
start
to
see
spatially
confined
signaling,
cues
and
markers,
which
means
that
it's
behaving
sort
of
like
an
embryo.
That's
that's
differentiating
into
different
parts.
A
So
these
gastroloids
are
just
these
mixtures
of
two
types
of
cell
they're
coming
together
and
then
they
start
behaving
like
a
real
embryo.
So
this
is
like
a
nice
example
of
what
you
know.
What
what
they're
actually
trying
to
do
here?
They're
trying
to
create
an
embryo
and
we're
gonna
actually
talk
about
another
paper
where
they
do
the
same
type
of
strategy,
where
the
embryo
isn't
necessarily
an
embryo
as
a
whole.
A
A
A
So
this
c
is
a
schematic
for
genetically
guided
engineering
using
stem
cells
engineered
with
inducible
genetic
circuits
to
pre-program
sulfates
associated
with
different
germ
layers.
So
this
one
here
is
where
they've
been
able
to
take
like
different
term
layers
in
the
embryo.
Combine
them
into
this
aggregate.
A
A
And
xenobots,
and
so
this
is
one
of
their
things
that
they're
doing
it's
getting
a
lot
of
attention,
especially
in
the
a
wife
community,
where
they're.
A
Is
an
artificial
ray
that
they've
created
using
a
3d,
elastomer
body
and
they're
able
to
generate
a
skeleton
and
then
muscle
just
they're
able
to
build
these
layers
of
tissue
on
top
of
this
artificial
scaffold
and
so
they're
able
to
assemble
this
artificial
ray
and
the
ray
can
actually
do
simple,
behaviors
by
using
optical
stimulation.
So
that's
this
technique
that
I
mentioned
before,
where
they
can
just
take
like
a
light
signal
at
a
certain
frequency
and
stimulate
the
cells
in
a
certain
way.
A
They
can
show
what's
being
expressed
in
the
cell,
but
they
can
also
make
the
cell
do
things,
so
they
can
make
the
cells
excited,
or
sometimes
they
can
drive
movement
of
cells
or
coordinated
movement
to
cells.
In
this
case,
I
think
they're
actually
able
to
move
the
ray
in
a
certain
direction
by
stimulating
the
cells
with
a
certain
light
frequency
and
then
they're
able
to
reduce
muscle,
contractions
and
move
it
forward
or
to
make
it
undulate
it's
its
rays
or
its
fins
here
and
so
they're
able
to
get
it
to
maneuver.
A
E
A
So
that's
a
lot
of
it's
a
lot
of
work.
I
think
this
artificial
ray
is
not
really
something
that
they've
created
it's
in
the
it's
in
the
planning
stages,
but
the
odds,
the
idea
of
what
they're
trying
to
do
here,
and
so
just
to
be
clear
on
that.
This
is
an
example
of
extracting
cells
from
xenopus
embryos,
allowing
these
cells
to
reassociate,
and
then
microscoping
removes
some
of
the
cells
and
that's
how
they
basically
try
to
form
some
of
these
xenobots.
A
So
this
is
so.
It
requires
a
lot
of
genetic
engineering
here,
they're
trying
to
engineer
gene
circuits
and
they're
trying
to
control
what
they
call
pluripotency,
which
is
this
ability
of
the
cell
to
shift
its
fate
or
to
make
fate
decisions
based
on
some
stimulus
that
they're
giving
it
so
cells
start
out.
Actually
the
embryo
starts
out
totally
potent
and
then
it
becomes
pluripotent,
and
then
you
have
pluripotency
where
you
have,
you
can
switch
between
different
sulfates.
A
So
if
you're
a
pluripotent
stem
cell,
you
have
a
restricted
set
of
fates,
but
you
can
change
those
fates
using
some
sort
of
signal.
So,
like
you
can
take
transcription
factors,
you
can
take
chemical
inducers
and
you
can
switch
the
state
of
the
cell
from
a
stem
cell
to
a
differentiated
cell.
Sometimes
you
can
go
from
a
differentiated
cell
to
something
that
looks
like
a
stem
cell,
but
this
is
how
they're
able
to
deal
with
a
lot.
You
know
how
they're
able
to
sort
of
manipulate
the
the
embryo
or
the
gastroid.
A
A
Okay,
surety
had
to
leave
early,
thank
you
and
jesse
had
to
leave
as
well,
but
we're
gonna
go
through
one
more
paper
here.
This
is
the
one
here
that
I
was
telling
you
about
where,
in
this
case,
they're
actually
arguing
in
a
sense
that
you
don't
need
to
have
a.
A
A
Areas
show
that
destruction
of
aggregates
of
mouse
embryonic
cells
with
an
experimentally,
engineered,
morphogen,
signaling
center
that
functions
as
an
organizer
results
in
the
development
of
embryo-like
entities
or
embryoids,
so
they're,
using
this
embryoid
method
again,
where
they're
trying
to
create
they're
taking
cells
or
putting
it
in
a
culture
and
they're
trying
to
create
something.
That's
like
an
embryo
that
has
the
capacity
to
sort
of
form
and
differentiate
like
an
embryo
but
they're,
actually,
cells
that
are
put
into
this
context
separately
from
an
embryo
in
c2,
hybridization,
immunolabeling
and
cell
tracking
transcriptomic
analysis.
A
So
these
are
all
methods.
Analysis
methods
show
that
these
embryoids
form
three
germ
layers
through
a
gastrulation
process
and
that
they
exhibit
a
wide
range
of
developmental
structures,
highly
similar
to
neuralis
stage
mouse
embryos.
A
So
these
are
basically
cell
cultures
with
a
bunch
of
stem
cells
or
where
pluripotent
cells
in
them,
and
then
they
can
sculpt
those
cultures
to
behave.
Sort
of
like
an
embryo
they're
able
to
coordinate
the
differentiation
patterns
there.
Okay,
so
embryonides
are
organized
around
an
axial
chloromisoderm
with
a
dorsal
neural
plate
that
displays
histological
properties
similar
to
the
murine
embryon
or
epithelium,
and
that
folds
into
a
neural
tube
pattern.
Anterior
posteriorly
from
the
posterior
midbrain
to
the
tip
of
the
tail.
A
So
there's
this
patterning
along
the
anterior
posterior
axis
in
this
case.
There's
this
they
get
this
sort
of
neural
tube
that
forms
and
it's
forming
from
sort
of
the
mid
middle
part
of
the
organism.
Maybe
the
midbrain,
which
is
sort
of
at
the
anterior
end
and
then
to
the
tip
of
the
tail,
which
is
the
posterior
end
lateral
to
the
coromizoderm.
Embryos
displays
semitic
and
intermediate
mesoderm
with
beating
cardiac
tissue
anteriorly
in
formation
of
a
vasculature
network.
A
Eventually,
embryoids
differ
differentiate,
a
primitive
gut
tube,
which
is
patterned,
both
anterior
posteriorly,
which
is
front
to
back
and
dorsal
ventrally,
which
is
top
to
bottom.
So
you
have
these
different
you're.
Getting
these
different
anatomical
axes
and
so
the
front
to
the
back
the
top
to
the
bottom
and
they're
actually
getting
these
things
that
look
like
a
primitive
gut
tube
and
a
neural
tube,
but
they're
but
they're,
not
you
know
I
mean
I.
D
A
Know
how,
if
they're
actually
going
on
to
form
they're,
not
going
to
necessarily
go
on
to
form
a
fully
formed
organism,
an
adult
organism
so.
C
A
A
You
can
take
a
bunch
of
stem-like
cells
and
put
them
into
these
organoid
form
these
organoid
structures
that
actually
form
or.
A
A
So
yeah,
one
limitation
that
may
explain
why
only
partial
development
is
obtained
in
vitro
may
be
related
to
the
method
used
to
induce
developmental
programs
in
embryonic
stem
cells
in
vertebrates.
The
signals
breaking
the
initial
spherical
symmetry
of
the
egg
and
instructing
cells
about
their
fate
and
behavior
are
present
in
the
egg
or
are
generated
by
the
cells
of
the
embryo
itself
or
by
its
extra
embryonic
tissue.
So
you
have
all
these
things
that
happen
in
the
mouse
embryo
you
have
a
primitive
streak:
extra
embryonic
ectoderm
and
visceral
endoderm
in
the
mouse
embryo.
A
A
A
A
B
You
heard
it
yeah.
The
hanging.
Droplet
is
very
simple.
You
take
a
cover
slip
and
put
it
you
use
it
with
a
microscope
slide
that
has
a
a
spherical
weld.
That's
been
carved
out
of
it.
So
there's
some
space.
Oh
right,
yeah,
you
drop
on
the
cover
slip,
a
drop
of
liquid
on
the
coverslip
and
you
turn
it
upside
down.
So
it
hangs
right.
A
So
this
is,
these:
are
the
formation
of
embryoids
by
instruction
of
vsc
aggregate,
so
the
morphogen
signaling
center.
So
here
they
have
these
small
aggregates
and
large
aggregates
they
induce
a
signaling
center
somewhere
in
the
organoid,
and
then
they
watch
the
you
know.
They
monitor
it
over
a
number
of
days
to
see
what
happens
and
it
turns
out
that
they
get
these
these
different
things
that
differentiate
with
a
signaling
center
involved.
A
So
this
is
position
over
time
of
mesoderm
cells
during
the
gastrulation
of
embryoids.
Maybe
this
is
a
better
figure
for
kind
of
what
we
want
to
talk
about
here.
So
this
shows,
I
guess
these
green
cells
are
the
these
are
broad
gfp
expressing
cells.
This
is
their
marker
for
the
mesoderm
cells
and
they're,
moving
towards
the
anterior
end
of
the
gastroid,
and
therefore
there's
sort
of
this
there's
sort
of
this
wave
over
time.
That's
driven
forward,
but
also
driven
sort
of
forward
by
wind
and
nodal
expression
in
the
back
end.
A
Here,
that's
driving
them
sort
of
forward,
and
then
they
come
back.
So
there
are
all
these
different
types
of
things
going
on
with
this.
So,
let's
see
c
is
an
interpreter,
drawing
mesoderm
cells
and
is
by
winthrop
nodal,
which
are
these
blue
arrows
here
in
the
signaling
center
were
observed
progressively
more
anteriorly
during
gastrulation.
A
And
so
this
is
this:
is
you
know
this
is
just
the
kind
of
stuff
they
can
do
with
this
kind
of
model
to
sort
of
get
a
handle
on
some
of
the
signaling
pathways
and
that
and
some
in
how
they
behave
in
these
aggregates
okay.
So
I'm
going
to
stop
there
and
I
think
that's
been,
I
think,
that's
all
for
today
in
terms
of
papers.
So
again,
if
you
have
anything
you
want
to
present
just
let
me
know
also.
A
I
have
some
papers
from
dick
that
he
sent
me
on
various
topics,
we'll
probably
go
through
some
of
those
in
the
next
few
weeks
and
if
it
might
not
keep
up
the
good
work
and
keep
presenting
on,
I
mean
we'll,
have
an
update
every
week.
I
think
on
your
work
about
the
same
length
that
you
gave
this
week
so
we'll
just
kind
of
keep
reviewing
your
work,
and
hopefully
you
stay
on
top
of
everything,
and
if
you
have
again,
let's
keep
in
touch
on
slack
all
week.
Make
sure
that
we
are
you
know.
B
All
right
bradley:
do
you
have
a
minute
after
we're
done
to
discuss
the
stuff
james
status,
yeah,
yeah,
okay,
so
thanks.
A
For
meeting
have
a
good
week.