►
From YouTube: DevoWorm #22: GSoC Project Updates, Smooth Moves in Diatoms, New Directions in C. elegans Modeling
Description
GSoC Coding period (week 1): Updates on D-GNNs and Digital Microspheres. Modeling cell and tissue tensegrity in COMSOL, review of Digital Bacillaria project and modeling movement smoothness, Papers on C. elegans positional mapping and lineage tree (embryogenetic) variation across Nematode taxonomy. Discussion about differentiation waves in the context of mosaic and regulative morphogenesis. Attendees: Susan Crawford-Young, Harikrishna Pillai, Karan Lohaan, Richard Gordon, Wataru Kawakami, and Bradly Alicea
B
C
D
D
Okay,
all
right
welcome
to
the
meeting.
I
know
yeah
hang
can't
be
with
us
today,
but
he
we've
been
talking
on
slack
quite
a
bit.
This
week
has
been.
I
forwarded
a
lot
of
that
stuff
to
wataru,
so
we
can
talk
about
some
of
that.
If
you
want
so
do
you
have
any
updates
from
anyone?
What's
the
latest
from
people.
E
E
So,
regarding
those
things
you
know,
especially
for
the
mass
shot
and
the
intergalactic
shot,
you
know
objectives.
So
I'm
thinking
you
know
something
more
hopeful
than
this
could
be.
You
know
incorporating
another,
but
neural
net,
like
the
solution
that
I
am
currently
working
at
is
somewhat
similar
to
a
linear
programming
solution.
Okay,
it's
like
we
have
these
like
the
sub
problems.
They
are
doing.
You
know
they're
solving
individual,
some
smaller
sub
problems,
and
then
you
know
they're
trying
to
recreate
the
3d
model.
E
So
you
know
like
the
dibr
volumetric
rendering,
where
you
know
you
have
a
3d
probabilistic
point
cloud.
Instead
of
you
know,
just
a
single
function,
giving
you
one
single
point
load.
So
it's
instead
of
just
a
single
point,
load
you'll
have
a
probabilistic
point.
Cloud
of
you
know
the
possible
shape
that
that
3d
model
could
have
so
you
know
this
could
probably
be
included
in
the
you
know.
Mash
shot
in
the
intergalactic
shot
portions
I'll
have
to
you
know,
make
it
more
concise
and
you
know
actually
check
what
the
key
objectives
for
doing.
E
E
D
B
E
For
now,
and
just
going
through
different
things,
you
know,
maybe,
if
with
regards
to
you,
know
more
like
guidance
towards
how
you
know
we'll
be
developing
the
final
model
I
think,
towards
when
we're
reaching
more
towards
the
you
know,
generating
the
3d
model
part
I
think,
then
I
might
need
more
guidance
as
to
you
know,
would
this
method
be
better
with
that
method?
I'll
just
be
showing
you,
you
know
outcomes
of
those
three
methods
that
I
will
be
using,
but
for
now
this
is.
E
So,
instead
of
like
generating
just
a
normal
point
cloud
using
the
current
algorithms,
it
will
just
be
relying
on
the
contour
data,
the
outlines
of
the
eight
images
that
are
there.
So
this
is
the
objective
method
that
you
know.
I
I
mentioned
in
my
proposal
that
I'll
be
using,
so
it's
just
generalizing
those
eight
outlines
and
then
it's
creating
a
point
cloud
based
on
those
eight
points.
So
when
we
use
a
neural
net,
I'm
thinking
of
using
the
whole
data
set.
E
E
The
outline
that
the
linear
programming
method
generates.
So
I'm
still,
you
know
trying
to
write
down
the
major
objectives
that
I'll
have
to
go
through
to
achieve
accuracy
using
that
method,
but
so
far
you
know
if,
if
there's
any
sort
of
improvement
that
can
happen,
it's
you're
probably
using
that
method.
Only
because
your
linear
program,
otherwise
will
should
give
you
some
results,
but
yeah.
E
F
F
E
E
D
F
D
H
D
D
D
D
D
Yeah,
okay,
yeah,
so
yeah
again,
you
know
make
sure
that
we
keep
in
contact
about
that
and
if
you
have
questions
or
if
you
have
like
thoughts
about
how
this
this
is
going
to
be
done.
Let's,
let's
talk
about
it.
I
don't
have
like.
I
don't
have
all
the
answers,
but
I
can
help
you
because
we've
done
this
a
number
of
times
so.
I
Yes,
so
this
week,
like
I
found
two
ways
in
which
I
think
I
could
proceed
with
and
one
of
the
ways
was
using
the
images
I
would
be
finding
the
camera
positions,
it's
a
method
called
exception,
and
with
that
I'm
using
those
positions,
I
can
background
to
create
a
point
cloud,
maybe
using
the
studio,
vision
package
and
that's
entirely
while
reading
more
about
it.
So
I'm
trying
the
cooking
with
that
method.
D
G
G
G
I
G
B
J
B
I
G
F
I
G
F
G
Yeah
yeah
we
have
a
book
coming
up
on
diatom
on
chain,
diatoms,
that's
diatoms,
that
form
colonies
and
chains,
and
that
of
course
includes
our
favorite
bacillaria
okay.
So
there
are
enormous
number
of
movies
of
basil
area
online
and
the
typical
movie
nowadays
is
made
at
30
frames
per.
Second,
that's
exactly.
G
G
Okay,
so
the
problem
is
this:
the
bacillaria,
if
you
have
a
single
diatom,
it
seems
to
move
in
a
jerky
fashion.
In
other
words,
it
moves
like
like
that.
Okay,
but
bastille
area
looks
like
it's
moving
smoothly,
so
the
question
is:
can
we
act?
We
need
to
actually
measure
that
from
movies
and
see
if
basilary
it
moves
in
a
smooth
or
jerky
fashion.
G
G
G
G
D
All
right,
yeah,
I
know
the
people,
the
gsox
students,
probably
haven't
seen
the
stuff
we've
done
in
vassal
area.
So,
oh
okay,
maybe
some
of
the
people
got
involved
before
then
yeah,
but
I
mean
it's
not
yeah.
It's
not
it's
not
that
hard.
It's!
Basically
there
are
these
organisms
called
diatoms
and
there
are
a
lot
of
different
diatoms
in
the
world
they're
marine
microorganisms,
and
they
have
very,
they
have
a
very
special.
D
You
know
they
have
a
phenotype,
that's
basically
an
accretion
of
silica
and
maybe
some
other
materials
they
grow
and
they
have
these
interesting
life
cycles
and
they
have
various
modes
of
movement
and
so
we're
interested
in
how
they
kind
of
move
around
in
colonies
and
even
in
single
cells.
So
this
this
is
something
that
you
know.
If
you're
interested
we
can
talk
more
about,
but
it's
a.
B
D
Very
big
area-
and
you
know
it's
it's
one
of
these
things
where
you
can
find
data
online.
We
started
that
project
by
going
to
youtube
and
finding
videos
that
people
have
made
just
kind
of
like
they
can
make
take
diatoms
put
them
in
a
like
a
liquid
culture.
You
know,
under
my
like
an
amateur
observer,
can
take
a
microscope
and
put
it
under
a
microscope
and
see
you
know
the
diatoms
moving
and
then
you
know
you
can
do
different
experiments
with
them.
D
D
D
D
G
B
D
Right
yeah,
so
yeah
for
people
who
don't
know
brownian
motion
brownie
in
motion
is
it's.
You
can
observe
it
if
you
put
like
bacteria
under
a
microscope,
sometimes
or
some
other
microorganisms,
and
they
kind
of
move
around,
they
dance
around
a
central
point.
They
kind
of
like
almost
like
they're
vibrating,
okay,
and
if
you,
if
you
look
at
it,
you
know
this
is
something
that
they
do.
They
kind
of
move
around
like
that,
and
sometimes
they
move
to
explore
their
environment
or
whatever.
D
But
people
have
characterized
this
with
some
sort
of
like
a
standard
distribution
like
a
gaussian
distribution
and
they
call
it
brownie
in
motion,
and
then
there
are
different
variations
on
that.
So
you
can
have
like
drift
where
it
kind
of
drifts
off
that
center
point
a
bit
and
it
kind
of
explores
outward
there's
also
what
they
call
super
diffusion,
which
is
where
it
goes
out
quite
a
bit.
D
So
it's
a
different
distribution,
now
you're
talking
about
like
an
exponential
distribution
where
it
starts
to
move
out
further
from
that
center
point
and
then
returning,
and
so
there
are
different
ways
that
these
these
types
of
biological
movement
happen.
I
know
we
in
my
other
group:
we've
talked
about
levee
flights,
which
are
actually
very
distinct
from
brownian
motion,
but
they're,
something
that
you
see,
often
in
foraging
where
the
organisms
explore
a
little
patch
and
then
move
to
another
patch
of
resources.
But
it's
like
this
long
distance
flight
every
so
often
that
they
make
okay.
G
D
Well,
yeah
diatoms
are
really
easy.
I
mean
compared
to
like
c
elegans.
Diatoms
are
fairly
easy
to
work
with
because
c
elegans,
you
know:
you're
you're,
tracking,
a
bunch
of
cells
and
they're
dividing
and
diatoms.
You
usually
have
like
a
single
cell
or
a
colony
of
cells
that
are
very
small,
smaller
number,
fairly
small
and
then.
G
D
We
often
look
at
the
the
edges
of
the
cell,
the
membrane
and
then
we
look
at
the
maybe
try
to
find
a
centroid
or,
if
we're
lucky,
we
can
find
like
a
nucleus
area,
and
you
know
use
that
as
as
landmarks
on
the
cell
in
basal
area,
for
example,
one
of
the
diatom
species,
you
have
different
landmarks
that
are
basically
the
same
from
cell
to
cell.
So
you
have
like
you
know
the
ends.
You
have
some
other
features
that
are
fairly
easy
to
identify.
D
F
D
B
Oh,
I
managed
to
draw
a
hexagonal
group
of
cells,
but
they
tend
to
fall
over
because
they're
not
held
up
but
they're
supposed
to
be
held
up
by
other
tissues.
So
I
have
to
figure
out
how
to
put
some
springs
on
the
corners
to
hold
them
up,
so
they
don't
get.
I
don't
get
empty
matrices.
What
was
the
other,
and
I
got
an
empty
matrix
error
and
too
many
eigenvectors
error
too
many
I
in
vectors?
Yes,
so
I
just
I'm
still
working
hard.
B
G
B
G
C
B
D
B
Yeah
yeah,
this
is
for
them
my
candidacy,
I
have
for
the
candidacy
in
my
department.
I
do
a
project
and
then
I
I
write
it
up.
It
has
to
be
four
thousand
to
six
thousand
words
and
it
has
to
be
related
to
the
phd
thesis,
but
not
the
phd
thesis.
So
I'm
not
sure
how
that
works,
because
I
might
want
to
use
it
yeah
for
something
else.
So
I'm
not
sure
anyway,
maybe
I'll
have
to
just
say.
Oh
if
you
don't
like
that,
I
have
this
ball
microscope.
D
B
D
D
Yeah
yeah,
but
yeah
they're,
just
you
know,
they're
just
short
videos.
It
doesn't
take
much
to
show
the
process
of
you
know
like
a
lot
of
times
like
in
basilari
in
the
colonies.
They
have
this
accordion-like
motion
where
you
want
to
capture
that
process
and
that's
pretty
much
the
video.
So
it's
not,
and
you
know
a
lot
of
the
videos
on
youtube.
D
They
have
a
lot
of
like
artifact,
like
you
know,
other
bacteria
in
the
water
and
things
like
that,
and
but
that's
something
that
can
be
filtered
out
or
really
it's
a
matter
of
finding
the
landmarks
and
tracking
them.
It's
not
that
difficult
with,
like
bacteria
generally,
don't
get
in
the
way
other
algaes,
but
sometimes
they
do
so.
It's
just
a
matter
of
like
yeah,
but.
G
G
D
G
Yeah
particle
image
lost
symmetry.
I
think
it
is
right.
We
tried
that
once
when
I
was
visiting
in
norfolk
virginia
and
the
guy
I
was
working
with
suggested.
We
did
it
on
a
single
diatom.
While
it
was
moving
through
a
bunch
of
particles,
and
he
suggested
that
it
looks
like
they
have
what's
called
anomalistic
normal.
G
In
other
words,
it
looked
like
a
diatom
might
actually
smooth
its
way
in
some
fashion
and
that
hasn't
been
explored.
He
doesn't
have
that
kind
of
equipment
anymore.
He
moved
and
it's
an
open
question
as
to
whether
that's
what
see
a
vascillaria
can
slip
through
water
more
easily
than
you'd
expect
from
stokes
law.
G
G
C
D
D
So
a
couple
of
things
I
found
this
week
that
are
interesting-
maybe
you
know,
especially
I
guess
they're
nematode
related,
so
the
first
one
I'm
going
to
talk
about
is
this
paper
and
I
posted
this
in
the
open
worm
slack,
which
is
there's
a
channel
called
recent
papers,
or
something
like
that-
and
this
is
a
group.
It's
max
tegmark
who's,
a
physicist,
ed
boren
who's
at
mit,
does
kind
of
like
these.
D
So
this
is
by
by
bmc
bioinformatics
in
it
just
recently
published
this
year.
So
what
they're
doing
in
this
paper
is
they're?
Actually
looking
at
cells,
they're,
looking
at
volumetric
images
and
they're
tagging,
nuclei
and
neurons,
and
they
want
to
like
pick
they
want
to
create
an
atlas.
That's
the
most
accurate
atlas
that
exists
so
open
worm
the
idea
of
building
a
worm.
D
You
know
a
digital
worm
comes
from
a
lot
of
the
initiatives
prior
to
that
where
they
were
collecting
data
on
the
cells,
because
in
c
elegans
you
can
identify
every
cell.
You
know
where
it
is
in
the
worm,
but
the
resolution
of
where
those
cells
are
is
hard
to
get.
I
mean
every
every
c.
Elegans
basically
has
the
same
sort
of
sulfate
map
in,
in
the
sense
that
you
know
where
every
cell
is.
D
D
D
D
Some
of
it
is
identifying
the
actual
cell
itself,
so
annotating
the
cell
properly,
and
so
in
this
paper,
they're
going
to
go
through
a
series
of
methods
that
they're
using,
which
is
this
neuropal
method,
to
build
a
more
accurate
map
in
this
paper
they
actually
do
reference
some
of
the
open
worm
data
that's
been
used
in
the
past,
and
the
open
arm
data
comes
from
a
number
of
different
groups
who
have
done
this
before
so
this
is
not.
You
know.
The
open
worm
group
hasn't
really.
D
You
know
collected
their
own
data,
been
using
data
from
other
groups.
This
is
a
new
sort
of
data
set,
that's
that
could
be
used
by
openworm.
If
we
had,
you
know
once
it's,
this
is
published
and
it
could
be
used
to
update
our
model
that
we
have.
So
the
background
is
so
they're
using
these
volumetric
is
a
tag
neuronal
nuclei,
and
this
is
something
that
people
do
in
neuroscience,
but
they
also
do
it
in
developmental
biology
and
in
biology
in
general,
using
different
markers
to
look
at
the
cell.
D
D
It's
it's
a
little
unclear,
sometimes
as
to
what
those
actually
are
in
the
cell,
but
you
can
use
those
to
tag
like
the
position
in
space
and
you
know
maybe
an
identity
of
the
cell
in
terms
of
where
it
is
relative
to
other
cells.
But
you
can
also
do
things
like
turn
on
other
types
of
fluorescent
markers.
So
sometimes
you
can
do
these
multi-channel
markers,
where
you
use
different
filters
on
your
microscope.
You
use
like
a
a
blue
channel
or
a
red
channel
or
a
green
channel
and
all
those
different
channels.
D
If
you
look
at
it
in
a
fluorescent
under
a
fluorescent
microscope,
you
can
see
different
markers
expressed
at
different
levels.
So
these
are
good
for
looking
at
different
things
that
determine
the
cell's
identity,
such
as,
if
it's
a
neuron,
if
it's
a
somatic
cell,
if
it's
some
type
of
neuron
and
so
forth,
and
so
you
can
tell
those
things
through
those
markers
and
you
can
combine
those
markers
in
different
ways.
D
So
there's
a
lot
of
information
that
can
be
gathered
from
just
microscopy
data,
not
just
these
shape
parameters
or
these
positional
parameters,
but
other
parameters
as
well
so
frequently
cell
identity
is
determined
by
aligning
and
matching
tags
to
an
atlas
of
labeled
neuronal
positions.
So
this
is
what
I'm
talking
about
with
respect
to
position
in
c
elegans,
for
example,
they've
assembled
a
lot
of
these
maps
using
the
neuronal
positions
in
the
worm,
so
they
know
kind
of
the
xyz
position
of
each
cell
and
they
know
where
it's
supposed
to
be
and
other
identifying
characteristics.
D
D
I
know
more
generally,
like
you
know,
doing
these
sort
of
transcriptomic
assays,
where
they
look
at
like
what
a
single
cell
was
expressing,
but
linking
that
to
some
sort
of
atlas
where
you're
able
to
generate
so
the
you
know
to
generate
a
map
of
these
different
properties
really
hasn't
been
done
in
c
elegans.
Now
the
the
allen
institute
in
seattle
they've
done
a
lot
of
that
with
human
and
mouse
brain
slices.
They've
actually
done
different
types
of
antibody
stains
and
different
types
of
put
different
types
of
fluorescent
markers
in
mouse
brains.
D
D
D
So
they're,
you
know
they're
trying
to
align
these
point
clouds
resulting
from
typical
cl
against
fluorescence
microscopy
data
sets.
So
we
have
these
data
sets
in
our
github
repository,
and
I
think
it's
this
one,
and
so
we
have,
of
course
the
data
sets
that
we
talked
about.
Last
week
we
have
cell
birth
and
death
timing
data.
We
have
raw
nucleus
data,
we
have
some
other
species,
so
there
there
are
different
data
sets
in
our
github.
D
There
are
also
data
sets
available
from
different
repositories
that
allow
you
to,
you
know,
get
the
get.
The
raw
data
get
something
that's
been
processed,
there's
this
alignment
process
that
needs
to
happen
as
well,
sometimes,
when
you're
working
from
something
that's
just
a
set
of
images,
I'm
not
sure
I
think
we're
doing
that
now
with
axolotl,
but
this
is
a
paper
if
you're
interested
in
some
of
the
state-of-the-art
techniques
people
are
using
for
c
elegans.
This
is
a
paper
to
read
and
go
through
a
little
bit
more
carefully.
D
Their
method
that
they're
introducing
in
this
paper
involves
a
tunable
learning
parameter
and
a
kernel
that
enforces
biologically
realistic
deformation.
So
they
use
a
algorithm
that
assumes
deformation
of
the
worm
and
they're
using
that
to
sort
of
correct
for
a
lot
of
the
noise
that
you'll
find
when
you're
trying
to
align
the
cells
with
each
other.
D
We
also
present
a
pipeline
for
creating
alignment
atlases
from
data
sets
of
the
recently
developed
neuropal
transgene.
So
this
is
a
new
method
that
they're
using
that's
a
transgenic
method
that
allows
you
to
generate
these
different
fluorescent
markers
and
get
them
to
express
in
different
cells,
so
they're
actually
doing
this.
D
The
secondary
part
of
this
is
creating
a
pipeline
for
these
different
methods.
So
you
have
these
different
methods
that
are
generating
data
and
you
want
to
align
them
into
a
an
atlas,
and
so
you
have
to
create
a
pipeline
for
that
in
combination.
These
advances
allow
us
to
label
neurons
in
volumetric
images
with
confidence
much
higher
than
previous
methods.
D
So
this
is
the
most
complete
full
body
c
elegans,
three
positional
neuron
atlas
incorporating
positional
variability,
so
we
have
some
variability
in
c
elegans
individuals.
Sometimes
it's
a
stochastic.
In
other
words,
you
know
they're
different,
tiny
changes
between
worms
in
terms
of
the
position,
because
you
know
just
that's
the
way
it
works,
but
sometimes
you
know
you
have
mutants.
D
You
have
other
types
of
situations
where
you
have
a
technical
error
and
so
accounting
for
all
those
things
actually
makes
us
much
much
more
accurate,
and
so
this
is
incorporating
positional
variability
derived
from
at
least
seven
animals
per
neuron
for
the
purposes
of
cell
type
identity
prediction
and
for
myriad
applications,
such
as
looking
at
cell
pha
maps
or
imaging
neuronal
activity
or
gene
expression.
So
people
would
be
using
this
atlas
to
look
for.
You
know
to
align
it
with
their
transcription
data,
for
example,
or
look
at
like
neuronal
activity.
D
There
are
different
types
of
ways
to
measure
neuronal
activity
and
to
map
it
to
the
spatial
map.
So
that's
what
they're
trying
to
do
here
so
in
this
paper
they
actually
go
through
the
method,
and
this
is
interesting.
This
figure
one
shows
examples
of
different
c
elegans
atlases
that
came
before,
and
so
here
the
blue
dots
in
this
atlas-
and
this
is,
I
believe
this
is
the
well.
This
is
the
the
anterior
end
of
c
of
the
c
elegans.
D
D
We
just
use
just
the
standard
set
of
units
which
could
be
mapped
to
millimeters,
so
they're,
using
these
cell
tracking
and
from
the
cell
tracking
information
about
the
units
that
are
derived
from
that
and
then
they're
scaling
it
to
the
appropriate
metric
measurement.
So
this
is
similar
to
what
we
have
for
c
elegans
in
the
embryo.
But
this
is
in
the
adult
and
the
blue
dots
are
yes.
G
D
G
G
D
These
are
the
red
dots
and
then
they're
they're
improved
upon
an
open
worm
because
they
use
some.
They
used
other
other
more
recent
data,
and
then
they
did
some
alignments,
and
so
these
are
these
represent
average
positions,
so
these
are
across
worms.
There's
going
to
be
some,
you
know
like
shrinkage
of
the
specimen
and
things
like
that
as
you're
measuring
it.
So
there
are
a
lot
of
things
that
come
into
play:
yeah,
okay,.
D
D
Yeah
yeah,
something
like
that
where
you
have
yeah:
okay,
okay,
so
this
is.
This
is
an
example
of
the
different
data
sets
that
exist,
and
this
is
an
example
of
like
they're
improving
upon
these
results.
So
you
can
see
that
there's
some
well!
You
can't
really
tell
from
here,
but
there's
you
know,
there's
there's
a
difference
between
the
two
and
yeah.
You
could
use
a
vector
diagram
to
see
the
difference
in
how
it's
what
it,
what
the
differences
are,
but
you
know
that
the
accuracy
is
not.
D
You
know
I
mean
you
can
see
like
in
the
white
data.
It's
like
following
these
edges
of
the
cube
here,
so
there
are
approximations
that,
are
you
know,
improved
you're,
just
improving
upon
some
approximations
and
that's
that's
what
you're
that's
what
they're
kind
of
improving
upon
and
then
they
show
their
neuropal
atlas
construction.
So
this
is
for
just
kind
of
cons,
reconstructing
data
from
these
you
know
aligning
the
neurons
using
the
labels
and
incorporating
unlabeled
neurons
into
this
framework.
D
So
they're
taking
you
know
these
labels
they're
using
that
information,
they're
aligning
the
neurons
and
then
they're
also
incorporating
unlabeled,
neurons
and
they're.
Using
this
neuropal
atlas
as
a
sort
of
a
source,
and
so
they
talk
about
some
of
their
pipeline,
they
talk
about.
I
don't
know
if
there
are
any
other
images
in
here
that
are
really
oh
here's
one
where
they
show
how
they're
modeling
the
phenotype
of
the
worm
so
they're
doing
this
sort
of
geometric
model
where
they
start
with
an
worm
image
at
the
top
left.
D
They
determine
the
hall,
which
is
the
top
center
images
black
image,
and
then
they
have
this
these
tangent
circles,
which
actually
map
out
sort
of
the
position
for
these
neurons.
So
the
neurons
you
find
them
in
the
head,
which
is
here
the
tail
and
then
they're
a
few
scattered
about
in
the
mid
center
of
the
organism
and
then
you're
trying
to
figure
out
like
the
how
to
approximate
that
position
best
they're
using
these,
these
tangent
circles,
1000
of
them
across
the
worms.
D
D
Then
this
is
a
sagittal
view
of
this
derived
atlas.
So
you
can
see
that
you
have
the
position
from
that
detail.
This
is
the
head.
This
is
the
tail
at
eight
hundred
nano
per
microns
ventral
dorsal.
You
have
this
variation
of
50
microns.
So
it's
it's
a
very
you
know
it
doesn't
have
a
lot
of
ventral
dorsal
depth
which
is
top
to
bottom
but
or
bottom
to
top,
but
there's
a
lot
of
length,
and
so
this
is
basically
showing
these
positions
and
so
they're
improving
upon
the
existing
data.
D
D
D
D
This
is
transverse
view,
so
this
is
just
looking
straight
down
from
the
from
the
front
of
the
worm,
so
it
doesn't
really
have
a
face,
but
it
would
be
the
face
of
it
and
you
look
down
the
worm
like
that,
and
you
see
this
transverse
view
just
kind
of
like
this.
If
you're
looking
kind
of
at
the
sort
of
the
front
of
the
brain,
I
guess
that
would
be
what
it
would
look
like.
D
So
these
point
clouds
can
be
used
like
this,
that
you
can.
You
can.
You
know
reconstruct
things
and
they
kind
of
look
like
the
real
thing,
but
they
just
kind
of
look
like
points.
That's
that's
kind
of
what
you
want
to
aim
for
in
your
point:
clouds
if
you're
looking
for
certain
subsystems.
So
if
you're
looking
for
a
brain,
you
know
it
would
be
like
all
the
neurons
and
they
should
be
arrayed
in
something
that
looks.
D
D
Yeah,
that's
actually
like
a
hole
in
the
in
the
connecto,
and
I
think
it's
just
like
there's
a
ring
of
there's
a
nerve
ring
in
here.
You
know
that
where
the
neurons
are
aligned
around
and
so
there's
an
area
where
there's
no,
you
know
the
connectome
is
a
hole
in
it.
Basically,
it's
the
center
part
of
the
worm.
D
I
D
So
this
is
where
you
have
your
original
points
in
your
distorted
points,
and
then
you
have
deformation
points
which
are
these
black
dots,
and
so
the
idea
is
you're
deforming
it
in
different
ways,
and
you
can
show
like
a
field,
a
deformation
field
that
moves
the
neurons.
So
these
are,
I
don't
know
why
they're
doing
this
necessarily.
This
is
for
hyperparameter
tuning
to
just
make
sure
that
this
method
is
robust
and
that
it
has.
D
You
know,
they're
able
to
test
it
in
different
ways
like
this.
It
can
deform
it
to
see
if
it
maintains
its
position
or
its
relative
position
to
other
individuals,
and
then
that's,
I
think,
that's
it
there's.
You
know
they're
just
doing
a
lot
of
testing
on
the
method
and
that's
the
paper.
It's
it's
a
pretty.
D
It
has
these.
You
know
this
sort
of
deterministic
development,
where
every
cell
in
the
lineage
tree
becomes
a
certain
type
of
cell,
and
we
have
like
this
sort
of
deterministic
type
of
developments.
We
have
linea
cell
lineages
that
we
have
in
a
b
lineage
that
has
a
variety
of
different
tissues
that
can
be
derived
from
it.
D
We
have
other
lineages,
such
as
d
and
e,
which
derive
specific
types
of
tissues,
like
muscle
and
and
epidermis,
and
and
then
a
germline,
and
so
you
know,
the
c
elegans
lineage
tree
is
very
simple
in
that
sense
that
you
have
this
type
of
development
where
it's
you
know,
and
it's
all
arranged
from
like
this
anterior
to
posterior
manner,
so
the
lineage
tree
actually
unfolds
according
to
some
anatomical
position.
So,
but
in
this
other
nematode
and
in
fact
nematodes
are
very
diverse
in
terms
of
their
development.
D
So
as
it
turns
out,
other
nematodes
have
different
types
of
development.
They
have,
you
know
different
amount,
different
numbers
of
cells
across
individuals
and
the
species
c
elegans
is
something
what
they
call
utilic,
which
means
that
it
has
the
same
number
of
cells
in
every
individual
in
the
species.
So
if
you,
unless
you
again,
you
have
mutants,
you
can
go
from
c
elegans
to
c
elegans.
The
males
actually
have
a
slightly
larger
number
of
cells.
D
But
again,
if
you
go
from
male
to
male,
you
have
the
same
number
of
cells,
but
that's
not
true
of
all
nematodes.
Some
nematodes
have
more
cells.
Some
have
fewer
some
have
a
little
bit
more
complex
developmental
state
stages
and
patterns,
and
then
you
get
to
this
nematode
here.
So
this
is
a
vastly
different
method
for
constructing
a
nematode,
which
just
means
that
its
development
is
vastly
different
from
what
we
find
in
c
elegans.
D
So
the
current
picture
of
embryonic
development
in
nematodes
is
essentially
shaped
by
c
elegans
and
its
close
relatives
as
their
pattern
of
embryogenesis
is
rather
similar.
It
is
often
considered
to
be
representative
for
the
tax
on
nematoda
as
a
whole,
but
of
course
nematode
is
a
very
diverse
taxon
and
includes,
I
think,
thousands
of
species.
D
Some
of
them
are,
you
know,
have
a
similar
type
of
development
to
c
elegans,
but
they
have
different
numbers
of
cells
like
I
said,
but
there
are
some
that
have
a
vastly
different
type
of
development,
so
here
we
give,
for
the
first
time
a
comprehensive
description
of
embryonic
development
and
an
ancestrally
diverged
nematode.
So
that
means
that
this
is
a
nematode,
that's
related
to
c
elegans,
very
far
back
in
evolutionary
history
and
that
there's
a
divergence.
D
So
these
are
all
three
different
things
that
are
different
between
these.
The
spatial
arrangement
is
interesting.
Cell
division
patterns
are
also
interesting
and,
of
course,
tissue
formation
is
interesting.
One
of
the
things
that
c
elegans
has
that
a
lot
of
other
organisms
don't
have.
Is
this
sort
of
deterministic
type
of
development
where
the
cells
will
become?
D
You
know,
its
fate
is
determined
sort
of
in
its
lineage,
whereas
a
lot
of
other
organisms
will
have
what
they
call
regulative
development,
where
there's
a
lot
of
cell
signaling
and
it
determines
what
precursor
cells
become
a
certain
tissue.
So
if
you
take,
you
know
any
number
of
cells
they
proliferate.
D
They're
in
a
certain
spatial
location,
they'll
become
that
tissue
and
c
elegans.
It's
it's.
You
know
the
tissues
are
related
to
their
spatial
position,
but
they're.
If
you
took
a
cell,
say
from
the
back
of
the
worm
in
the
in
the
posterior
end
and
moved
it
to
the
anterior
end.
That
cell
would
not
transform
into
anything
like
the
neighboring
cells,
whereas,
if
you
did
this
in
another
organism
like
a
mouse,
you
would
that
cell
would
take
on
the
fate
of
where
it
was
placed.
D
So
you
know
if
you
put
took
a
muscle
cell
out
of
the
posterior
and
out
of
the
tail
of
a
mouse,
and
you
put
it
in
the
brain
it
might
take
on
a
neuronal
fate.
Of
course,
it
also
might
become
a
cancerous
cell,
but
that's
a
different
topic.
So
this
is
these.
These
two
different
species
vary
in
different
ways.
D
D
The
third
difference
is
construction
of
major
parts
of
the
body
from
duplicating
building
blocks,
consisting
of
rings
of
cells,
a
pattern
showing
some
resemblance
to
segmentation.
So
this
is
kind
of
in
c
elegans.
We
don't
really
see
segmentation,
we
see
like
nerve
rings
and
things
like
that,
but
in
these
nematodes
they
have
rings
that
are
replicated
across
the
body
from
anterior
to
posterior,
and
so
it
ends
up
looking
like
a
segmented
worm.
D
D
So
that's
interesting.
A
fourth
difference
is
prominent
differences
in
sulfate
assignment,
which
can
be
best
explained,
with
a
global
shift
affecting
all
somatic
founder
cells.
So
this
is
just
you
know,
the
idea
that
you
have
these
differences
and
how
sulfate
is
assigned
between
c
elegans
and
this
other
species-
and
you
know
they
don't.
It
just
goes
back
to
like
a
lot
of
the
early
founder
cells
of
the
lineages.
So
in
c
elegans
you
have
eight
founder
cells
at
the
eight
cell
stage,
and
these
each
these
founder
cells
found
a
lineage.
D
D
D
This
species
that
they're
interested
in
or
this
what
they
call
clade,
which
is
a
group
of
species
that
they're
interested
in
and
c
elegans,
is
way
up
here
at
clade,
nine,
so
c
elegans
is
diff
is
separated
by
quite
a
bit
of
evolutionary
distance
and
they
share
a
common
ancestor
right
about
here,
which
is
this
way
back
in
time.
It's
almost
to
the
base
of
all
nematodes.
D
D
They
call
it
p1
and
s1,
and
then
s
and
p
so
p
is
here.
This
is
the
usually
the
posterior
end
of
c
elegans.
S1
corresponds
to
a
b
and
s1,
of
course
divides
into
three
different
cells
here
and
then
there's
this
whole.
You
know
set
of
s
lineages
in
this
other
nematode
that
don't
necessarily
mirror
what's
happening
in
c
elegans.
There
are
differences
in
how
these
lineages
emerge,
and
so
you
can
see
that
there's
actually
some
asymmetries
here,
because
you
start
with
three
cells.
Instead
of
you
know
always
dividing
by
two.
D
You
know
one
to
two
cells.
You
get
these
three
three
s
cells
and
then
one
p
cell.
So
that's
an
asymmetry
that
you
see
very
early
on,
and
so
then
that
has
an
effect
on
the
entire
lineage
tree
and
how
that
unfolds,
and
then
they
have
some
other
here.
They
have
some
examples
of
different
spindles
in
the
cell,
so
this
isn't
necessarily
very
valuable,
at
least
at
the
sur
on
the
surface.
D
But
if
you're
interested
you
can
read
about
that
more
and
then
these
are
the
lineage
and
fate
maps
here
for
the
difference
species.
So
this
is
showing
the
lineage
tree
here
of
c
elegans,
the
lineage
tree
of
this
new
species
and
then
a
nanus
which
is
another
nematode
that
is
yet
different,
but
it's
closer
to
c
elegans,
and
so
here's
the
lineage
tree
here
where
you
see
these
dip.
So
this
is
the
lineage
tree
here.
This
is
the
fate
map
and
then
these
are
the
different
differences
in
them.
D
So
you
can
see
that
there's
quite
a
bit
of
difference
between
the
species
and
c
elegans
and
then
ananis
is
a
little
bit
different.
It's
actually
closer
to
what
you
see
in
c
elegans,
and
so
this
is.
This
is
a
cleavage
pattern
and
fate
designation
here.
So
this
is
reconstructions
of
early
embryonic
stages
based
on
4d
microscopy,
so
these
are
actually
in
their
position
in
space.
D
So
these
are
cells.
These
are
sister
cells,
where
these
lines
are
connecting
the
sister
cells,
the
products
of
a
cell
division,
and
so
you
can
see
that
in
in
the
c
elegans
and
then
this
other
species,
you
see
the
differences
here,
so
you
can
see
these.
These
are
what
they
look
like
in
their
spatial
positions
and
then
they
show
these
lines
between
them,
which
are
these
lines
that
connect
the
related
cells
and
just
showing
their
spatial
location,
and
so
that's
some
interesting
work
and
then
formation
of
rings
of
cells
and
intercalation.
D
So
you
see
that
you
have
these
rings.
That
form
in
a
is
a
is
the
hypodermis
b
is
where
you
see,
hypodermis,
transforming
and
then
c
is
where
you
have
this.
I
can't
I
don't.
I
don't
see
where
this
is,
but
this
is
just
this
process
going
through
and
you
can
see
these
rings
that
form
in
the
embryo
here
of
these
cells.
So
this
is.
These
are
two
microscopy
images,
and
this
is
a
model
here,
prime,
and
so
you
can
see
that
they
form
these
rings
that
are
repeated
and
so
yeah.
D
They're
really
nice
reconstructions
and
they
show
these
colored
beads
here,
showing
lineage
membership
fate
and
this
they
do
a
section
across
them
down
like
the
transverse
axis,
like
we
saw
with
the
adult
c
elegans
connectome
or
a
set
of
neurons,
where
you
see
sort
of
looking
down
from
the
head,
and
so
this
is
a
nice
paper.
It
kind
of
gives
a
different
perspective
on
development
from
you
know,
looking
at
different
embryos
in
the
same
in
the
same
taxonomic
group,
okay,
so
hari
krishna
had
to
leave.
Thank
you
for
attending
hari
krishna.
D
G
G
I
speculated
in
this
old
book
of
mine
that
the
difference
between
mosaic
organisms,
which
is
c
elegans
and
regulating
organisms
like
mammals,
is
the
range
of
the
differentiation
wave,
in
other
words
in
c
elegans,
something
corresponding
to
the
differentiation
wave
would
be
confined
to
a
single
cell
and
in
mammals
it's
confined
to
a
tissue
which
could
be
millions
of
cells,
okay
and
and
the
differentiation
waves
make
the
difference.
So
it's
a
diff
difference
between
not
going
beyond
one
cell
or
going
to
n
cells
right.
G
Okay,
so
the
question
is
in
this
thing
seems
seem
a
little
backwards
here,
but
the
question
is:
is:
is
this
all
this
new
nematode
that
we're
looking
at
today?
Does
it
regulate
yeah?
Is
it
a
regulating
one
and
that
may
account
for
the
difference?
The
way
it's
backwards
is
that
it's
phylogenetically
earlier
than
c
elegans,
which
is
what
I
didn't
expect.
J
D
G
G
Okay,
so
in
any
case,
it'd
be
nice
to
see
if,
if
there's
any
way
of
relating
pretending
that
their
differentiation
waves,
which
they
don't
examine
and
see
if
see
if
their
data
would
fit
that
kind
of
speculation
or
not
that
that
these
are
more
like
regulating.
Now
I
remember
in
the
old
literature
on
snails.