►
From YouTube: DevoWorm #28: Computation of Development/Regeneration in Single-Cell Models, Open-source Initiatives
Description
GSoC Coding period (week 9 and 10): Updates on Digital Microspheres and D-GNNs. Review of activity on the Devolearn Github organization. A review of computing Diatoms (Digital Bacillaria). Overview of NetSci 2022. Review of the transcriptional program for regeneration in Stentor. Attendees: Richard Gordon, Jiahang Li, Morgan Hough, Alon Samuel, Joshua Brewster, and Bradly Alicea
A
B
To
meet
you,
so
I
I
was
waiting
for
a
few
more
people.
I
know
you
tucked
in
the
slack,
but
maybe
you
could
introduce
yourself
that'd
be
good.
Oh.
C
Yes,
yeah
yeah,
that
would
be
nice
yeah
and
yeah
and
see
people.
Yes,
I
did
write
some
some
stuff
on
slack,
but
I
felt
like
oh.
C
That's
that's
nice
and
yeah
happy
to
kind
of
like
to
join
yeah,
really
new
to
kind
of
like
everything.
So
we
still
like
reading.
B
Yeah
yeah
reading
over
everything
yeah
it's
a
lot
to
get
into,
but
well
yeah.
I
don't
know
we
usually
have
our
gsoc
updates.
If
gsat
people
make
the
meeting,
you
know
we'll
get
into
that,
and
I
have
some
things
to
talk
about
so.
C
The
last
one
yeah
helped
me
like
possibly
go
into
the
it
helped
me
like
that.
You
posted
it
on
on
youtube.
Yeah
yeah,.
B
C
B
All
right,
well
hi
hi!
So
how.
D
D
Yeah
I'm
good
yeah,
yeah
and
yeah,
and
I'm
not
sure
if
long
we
will
compete
because
since
then
he
has
some
updates
regarding
the
step
one
of
our
gso
projects,
yes,
and
based
on
his
updates,
I
will
make
more
updates
about
the
step
two
of
our
projects
in
this
week.
I
guess-
and
besides
I
have
no
other
updates.
Yeah.
B
All
right,
you
don't
have
any
more
updates.
How
are
things
going
with
your
project
like?
How
are
things
going?
I
don't
know
like
have
you
heard
from?
Have
you
been
talking
with
otaru
on
a
regular
basis
and
getting?
I
know
that
step
one?
You
don't
have
the
update
for
that,
but.
D
D
I
mean,
really
implement
the
graph
neural
network
for
cell
tracking
paper
as
our
gso
projects,
yes,
and
with
charlie
and
longhuit.
They
are
working
on
the
step,
one
of
the
gm
fossil
tracking
and
I've
discussed
with
longhue
several
times.
Last
weekend.
D
She
has
successfully
implemented
the
cause
of
the
original
author,
but
I
don't
know
where
he
is
where
he
works
regarding
the
gsu
projects.
Exactly
so
perhaps
if
he
will
not
come
to
our
meeting
today,
perhaps
I
would
discuss
with
long
way
later.
Okay
in
this
week,
yeah
yeah
that.
D
Good
yeah
and
besides
yeah
yeah
and
in
this
week,
if
I
see
that
longhui,
has
made
a
lot
of
updates
in
the
last
week.
So
perhaps
in
this
way
we
can
make
more
progress
regarding
the
second
step,
which
is
to
convert
the
outputs
of
step
one
into
the
graph
structure
and
to
implement
the
cell
tracking
model,
which
is
actually
the
step
three.
D
B
C
B
So
we
have
a
couple
more
weeks
of
the
project
and
if
you
do
need
an
extension,
let
me
know
I
can
request
it,
but
our
target
is,
I
think,
the
first
or
second
week
of
september
off
to
consult
the
calendar
so
yeah.
I
think
I
think
if
you
start
two
and
three
now
and
get
going
on
that,
and
I
think
it
should
all
get,
you
should
be
able
to
get
there
by
the
deadline
and
and
have
something
pretty
good.
So.
B
Welcome
dick
and
morgan.
B
Elon,
would
you
like
to
introduce
yourself.
C
Yes,
yeah
I'd
love
to
happy
to
meet
all
of
you,
hi
yeah,
just
joined
and
kind
of
like
because
I
want
to
contribute
some
of
like
my
skills
and
time
to
kind
of,
like
maybe
bio,
research,
and
I
saw
that
the
open
world
is
kind
of
like
a
good,
like
maybe
opportunity
for
me
to
do
that.
C
I
have
like
I'm
working
as
a
data
scientist
in
nlp
at
the
moment,
but
I
have
experience
in
computer
vision,
both
in
the
neural
network,
cnns
and
other
reinforcement,
learnings,
algorithms,
and
also
maybe
classic
image,
processing,
video
algorithms-
and
you
know
previous-
and
not
in
this
job,
but
in
my
previous
one
and
yeah
my
kind
of
like
education
is
that
finished
electrical
engineering,
2012
image
processing
and
what
was
it.
C
I
think
it
was
electro
optics,
yeah
focused
on
that
and
yeah
I'm
living
in
bristol,
originally
from
israel
and
yeah.
That's
kinda
like
that's
me,.
B
Oh
good
yeah,
that's
great,
so
yeah.
We
have
a
lot
of
things
that
we
do.
I
don't
know
it's
like
it's
hard
to
really
kind
of
hook
into
something,
but
if
you
attend
the
meetings
and
watch
kind
of
some
of
the
things
we
talk
about,
it's
like
if
you
find
something
you
like
or
if
there's
something
that
you
see
that's
kind
of
been
put
aside,
you
might
bring
that
up
and
ask
about
it.
Other
than
that.
You
know.
It's
kind
of
you
know
it's
hard
to
kind
of
slot
people
into
a
place.
C
Yeah
good
yeah,
that
sounds
sounds
great
kind
of
like
yeah.
I'm
reading
a
lot
like
also
like
watching
some
stuff
on
biology
that
I
kind
of
like
all
this
morphogenesis
and
chromosomes
and
dna.
I
need
to
kind
of
like
just
yeah
the
background
from
that
so
yeah.
Whatever
I
can
yeah,
I
can
help
I'm
gonna
check
and
then
yeah
hop
on
to
that.
That's
gonna
be
great.
B
That's
great
yeah:
what's
one
of
the
things
we
do
in
the
group?
Is
we
kind
of
bridge
biology
and
computer
science
or
traditional
computer
science,
and
so
that
that's
a
good
opportunity
to
get
into
that,
like
you
know,
cross
disciplinary
area
so.
E
Alan,
what's
your
experience
in
image,
processing.
C
Yes,
I
did
a
few
projects
with
tracking
in
video
tracking,
basically,
and
also
some
more
of
like
maybe
simple,
tracking
related
to
neural
networks.
Yeah,
like
detection,
neural
networks
and.
C
What
we
had
a
project
that
we're
tracking
just
like
people
to
help
them
like
monitor?
It
was
like
a
safety
project
for,
like
you,
know
the
kids
to
kind
of
like
say:
they're,
not
gonna,
be
left
alone
in
the
backyard.
So,
for
example,.
A
C
I
was
kind
of
like
working
on
yeah
on
implementing
like
a
neural
network.
It
was
mean
that
basically,
it
means
to
take
something,
that's
kind
of
out
there
and
then
train
it
on
our
like
data
and
then
in
just
on
a
client,
nice.
E
Okay,
but
we
have
that's
interesting,
we
have
a
tracking
problem
that
we're
looking
for
somebody
to
get
interested
in.
Oh,
that's,
that's
cool!
The
the
problem
is
it's
a
microscopic
algae
called
diatoms
and
it's
a
particular
species
called
bacillaria
paradoxes:
okay
and
the
cells
slide
against
each
other
out
to
the
ends
of
the
cells
and
then
back
okay,
and
you
want
to
track
the
cells
want
to
track
the
cells.
E
There
are
probably
hundreds
of
movies
already
online
that
one
could
work
with,
and
the
outstanding
problem
right
now
is
whether
they
move
in
a
smooth
manner
or
whether
they
move
in
a
jerky
matter.
E
Okay,
so
we.
E
No,
no,
no,
we
don't
know
which
movement
they
they
actually
don't
know.
The
problem
is
that
they
move
fast
enough,
so
that
your
eye
can
see
them
as
moving
smoothly,
even
though
it
might
be
jerky,
but
typical
movies
online
would
be
about
30
frames
per
second,
okay
yeah,
okay,
nice.
We
we
have
tracked
individual
diatoms
and
their
movement
is
jerky,
and
people
have
looked
at
these
and
said.
E
Oh,
it
looks
smooth
to
me,
but
I
don't
believe
I
don't
believe
people
okay,
so
we
need
a
definitive
decision
as
to
whether
the
movement
is
smooth
or
jerky
and
nice
okay.
So
the
idea
is
to
track
the
individual
cells,
look
at
their
relative
motions
and
see
if
the
motion
over
time
is
best
fit
by
a
smooth
curve
or
best
fit
by
a
jerky
motion.
E
Right
right,
smooth
would
mean
like
a
car,
a
car
will,
you
know,
take
pictures
of
the
car,
no
matter
what
it's
doing,
even
if
it's
accelerating
the
motion
is
smooth
okay.
But
if
you
take
I'm
trying
to
think
if
you
take
pictures
of
a
moth
flying
they're
deliberately
jerky,
because
they're
trying
to
avoid
being
eaten.
C
E
A
E
A
E
C
Sorry,
what's
an
algae,
oh
lj,.
E
E
E
B
B
The
cells
and
finding
the
and
refining
the
boundaries
of
the
cells,
but
you
know,
there's
a
lot
in
the
motion
that
needs.
You
know
that
needs
to
be
optimized
or
or
kind
of
figured
out.
So
you
know
you
can
get
you
have
to
the
first
step
and
we
kind
of
have
the
code
is
to
define
the
cell
and
the
with
against
the
background.
So
it's
not
as.
D
B
It
sounds
because
there's
a
lot
of
like
different
things
in
the
water
that
you
know
when
you
take
an
image
of
it,
that
kind
of
obscure
the
cells
sometimes
and
then,
of
course,
they're
moving
at
a
certain
rate.
So
you
know
you
could
take
a
still
image
frame
of
that
movie
and
you
can
see
it.
You
know
it's
it's
a
static
thing,
but
then
you
have
to
stitch
them
together
in
terms
of
the
movement.
B
So
is
the
cell
in
this
frame
the
same
as
the
cell
on
the
next
frame,
because
it's
moved
and
then
you
know
it
kind
of
moves
back
and
forth.
So
you
have
to
pick
up
that
oscillation
over
frames
and
then
the
boundary
of
the
cell
isn't
always
very
clear,
and
this
is
a
problem
with
a
lot
of
our
work
that
we
do
with
cell
segmentation
is
that
the
boundaries
of
the
cells
are
often
very
clear
in
some
images.
E
C
I
think
it's
good
that
I
think
I'll
start
like
to
using
with
a
data
science
project
to
look
at
the
data
and
then
like
see
like
yeah,
exactly
how
it
looks,
as
you
say
like
if
it's
kind
of
like
it's
easy
to
segment
from
the
background,
is
to
track
with
also
like
yeah.
They
also
have
simple
algorithms,
like
optical
flow,
something.
E
E
E
E
A
C
When
you
say
like
track
so
you're
tracking
them
kind
of
like
this
open,
it's
kind
of
it's
all
like
a
one.
One
thing
to
track
like
this:
no,
you
attract.
E
The
individual
cells
yeah
uh-huh,
the
colony,
will
be
from
a
few
cells
to
hundreds,
okay,
okay,
so
imagine
cards
of
a
few,
a
few
cards
or
a
large
number
of
guards,
and
each
car.
E
E
E
Yes,
okay,
if
we
can
find
movements
or
make
movies,
we
have
at
least
one
collaborator
in
germany
who
makes
movies
of
these
guys.
Oh
that's
good,
okay,
yeah,
but
there's
so
many
movies
online.
I
suspect,
there's
sufficient
to
resolve
the
question
of
smooth
versus
jerky
movement.
A
C
E
E
Well,
those
are
probably
what
are
called
they're
all
kinetic,
diatoms,
okay,
so
those
are
most
diatoms
are
individual.
This
is
a
strange
colony
and
it's
been
named.
It's
been
given
about
three
different
names
in
terms
of
species,
but
it's
where
it's
worldwide
in
distribution
and
it
likes
especially
brackish
water.
C
E
E
We've
shown
this
we've
shown
that
just
turning
the
lights
off
is
sufficient
to
make
trigger
that
behavior
turning
it
on
and
they
go
back
to
the
back
and
forth
motion.
B
There,
it
is,
I
think,
it's
yeah,
an
issue
with
is,
I
think,
that's
fun.
A
B
Yeah,
do
that:
okay,
yeah,
there's
someone's
email
and
yeah.
I
don't
know,
what's
happening
with
morgan's
mic,
but
yeah
it's
I
don't
know
if
you
want
to
put
it
in
the
chat,
yeah
or
you're
gonna,
say:
oh
he's
going
to
go
out
and
come
back
in
okay,
okay,
so
I
got
an
update
from
hari,
krishna
and
and
quran
last
week
and
their
projects
are
going
well.
B
I
mean
they're
they're
working
on
some
they're
kind
of
at
the
phase
where
they're
trying
to
map
everything
to
the
sphere
and
so
now
they're
getting
that
they're
working
on
that
and
we're
excited
about
the
outcome
of
this
because
it's
it's
kind
of
like
two
complementary
approaches,
the
same
problem.
How
do
you
take
microscopy
images
captured
of
an
embryo?
B
So
we
can
ask
a
bunch
of
questions
about
that
and
oh,
are
you
back
morgan?
A
B
A
All
those
videos
are
from
yeah
yeah,
but
it's
probably
good
that
I
figured
out
what
my
what
my
problem
was
here:
okay,
cool
yeah
but
yeah,
and
then
I
I
had
a
link
from
ucl
on
on
these.
A
Because
it's
it's
they're,
not
just
algae,
but
they've
also
got
this.
This
like
is
it
silicon
based
yeah,
yeah,
they're,
glass,.
A
Yeah
I
mean
yeah
and
and
at
the
exploratorium
here
in
san
francisco,
they've
got
they've
got
an
exhibit
of
kind
of
like
all
the
different
silicon
shells
and
scallops
and
the
yeah
just
the
the
fascinating,
complex
structures
and
and
diversity.
You
know
is
that
online
at
all
I'll
check
I'll
check,
I'm
definitely
going
to
be
going
down
more
to
exploratorium
and
and
seeing
you
know
just
because
they've
got
so
many
resources
for
for
teachers,
and
you
know,
educators,
and
I.
A
Let
me
so
I
I
switched
computers
so
give
me
a
second,
but
I've
got
this.
This
good
link,
but
yeah
exploratorium
has
you
know,
has
live
microscopes
right.
A
It's
it's
a
great
spot.
A
We'll
see,
I'm
definitely
yeah
yeah
yeah
I'll,
go
down
and
take
more
pictures.
Okay,.
B
Great,
yes,
so
yeah
speaking
of
the
gsoc
project
and
the
what
we
call
the
digital
microsphere,
which
is
a
spherical
atlas
or
whatever
it's
going
to
look
like.
I
did
propose
that
we
do
a
paper
on
this,
so
it's
going
to
be
myself
dick
and
a
hare
krishna
and
quran
and
it'll
sort
of
sketch
out,
I
guess,
like
yeah
and
susan
okay,
and
so
it's
going
to
sketch
out
some
of
the
things
that
we've
been
talking
about,
like
the
methods
for
creating
these.
B
You
know
having
that
all
in
a
paper
talking
about
like
the
sort
of
the
significance
of
it.
B
So
the
idea
is
that
you
want
to
have
this
spherical
mapping
of
this
embryo
now,
just
because
the
embryo
is
spherical,
we
want
to
have
like
a
geometric
or
a
topological
model
of
it,
but
because
you
can
do
things
with
it,
you
can
actually
see
things
in
different
ways
and
do
analyses
on
it
hopefully,
and
we
can
get
a
new
understand
sort
of
a
novel
understanding
of
embryogenesis
by
having
that
sort
of
you
know
doing
things
on
a
surface
like
that.
B
So
you
know
there
are
a
number
of
questions
that
we
can
ask
here.
Can
we,
if
we
once
we,
what
they
call
montaging
the
images
which
is
putting
them
on
the
sphere
and
then
representing
a
time-lapse
sequence?
We
can
do
the
following,
and
so
there
are
a
number
like
seven
different
potential
things.
We
can
do
look
at
like
different
aspects
of
straight
local
strain
and
and
propagation
across
the
surface
of
waves.
B
The
different
loci
of
initiation
sites
of
we
call
differentiation
waves,
which
sort
of
kickstart
a
differentiation
process
of
cells
within
the
region
that
the
wave
crosses,
whether
those
places
show
any
signs
of
altered,
mechanical
stress
or
strain
ascertain
any
rules
by
which
termination
of
one
wave
triggers
the
next.
B
Then
you
know
other
things
you
can
take
out.
You
can
ablate
cells
and
see
what
the
effects
are
understand.
The
relationships
between
what
what
they
call
morphogens,
which
are
these
theoretical
molecules
that
drive
morphogenesis,
at
least
theoretically,
that's
what's
predicted
in
differentiation
waves,
which
is
this
other
theoretical
concept
that
we
can
test
more
accurately
in
this.
This
type
of
environment.
B
B
Yeah
yeah
generalized
montaging
to
embryo
shapes
when
deviating
from
spherical,
which
is
what
does
that
mean.
E
B
Yeah
and
then
coral
a
differentiation
waves
with
electrical
effects,
so
this
is
something
that
we've
talked
about
with
respect
to
some
of
the
stuff
michael
levin
does,
with
looking
at
electro
bioelectricity
in
in
development,
but
also
correlating
with
other
things
in
the
in
the
developmental
process.
E
E
So
and
nobody's
ever
figured
out
what
that's
for.
B
Yeah,
so
that's
yeah,
those
are
like
the
seven.
I
guess
things
that
we
can
explore
and
you
know
those
will
get
refined
as
we
go
along
and
then
we
have
the
references
which
are
you
know
just
right
now,
just
kind
of
things
that
we
thought
about
that
are
directly
relevant
to
this,
but
just
be
you
know,
I
guess
we'll
be
developing
this
paper
as
we
go
along,
especially
once
the
projects
are
completed,
we'll
see
what
we
can
do
and
you
know
we'll
have
the
methods
worked
out.
B
This
is
why
I
recommended
to
the
gsox
students
that
you
have
good
documentation
for
your
projects,
because
we're
probably
going
to
be
writing
papers
on
this
or
things
are
going
to
come
out
of
this
that
are
going
to
be
valuable
later
on.
So
definitely,
if
you're
doing
your
gsog
project,
you
know
this
is
the
time
to
kind
of
think
about
or
kind
of
put
the
documentation
out
there.
B
I
think
we
were
talking
jiahan
and
we're
talking
about
the
paper
that
there
he's
trying
to
reimplement
for
the
graph
neural
network
or
that
I
guess
the
cell
tracking,
and
so
you
know,
if
you
have
that
documentation,
I
mean
there's
a
lot
of
missing
parts
there
in
the
paper.
But
in
the
paper
you
know
you
want
to
have
a
good
description
and
then
you
want
to
be
able
to
link
to
the
documentation.
So
it's
definitely
something
that
we
want
to
that.
B
B
Yeah
yeah,
if
people
are
interested
in
contributing
to
this
yeah,
please
let
us
know
I
mean
you
know
it's
it's
a
matter
of
going
through
the
the
once
we
get
the
software
in
place.
The
software
is
going
to
be
made
public
the
open
source
so
we'll
have
some
sort
of
portal
for
people
to
use
it.
B
So
you
know
you'll
have
like
you
know,
maybe,
like
you
know,
you'll
go
to
the
website
that
we
have
and
you'll
be
able
to
interact
with
it
and
then
maybe
download
it
and
run
it
locally
on
your
machine,
and
then
you
know
we'll
be
wanting
to
test
different
things
with
it.
So
we'll
see
what
we
can
actually
do
once
once
the
software
is,
you
know,
ready
and
of
course
I
think,
krishna
and
karan
will
keep
working
on
this
after
the
project.
B
You
know
for
reason
purposes
of
the
paper.
Maybe
we
can
get
some
additional
functionality,
but
this
is
going
to
be
like
a
process
of
once
we
get
the
once
the
gsoc
project
ends.
We'll
have
this
nice
product
we'll
you
know,
have
it
as
an
open
source
tool.
People
will
start
using
it
hopefully,
and
then
we'll
also
be
working
on
this
paper
working
on
some
things
in
the
group
testing
different
ideas.
So
definitely
you
know
stay
tuned
if
you're
interested
in
getting
involved
in
this.
B
Yeah
yeah,
okay,
so
then.
C
C
B
Kind
of
that's
kind
of
an
overview
of
the
whole
group.
You
know
the
activities
and
it's
kind
of
out
of
date.
I'm
gonna
have
to
update
it
a
bit,
but
we
have
like
for
our
different
software
and
endeavors.
We
have
different
documentation
so
for
the
diva
learn
project,
which
is
a
lot
of
the
stuff
we
do
with
graph
neural
networks
and
pre-trained
models.
We
have
documentation
for
that.
B
We
have
documentation
which
I'll
send
you
for
the
digital
basil
area,
which
is
the
thing
we
just
discussed,
and
there
are
other
areas
where
it's
like
sub
areas,
but
yeah
I
gotta
I
get
to
update
the
read
the
docs
for
open
worm.
It's
been
a
while
and.
C
B
A
B
So
speaking
of
diva
learn,
I
wanted
to
go
over
what's
going
on
in
the
diva
learn
organization
on
github,
so
this
is.
This
is
the
product
of
like
a
couple
of
years
of
gsoc
projects
and
we've
brought
this
to
sort
of
an
open
source
platform.
That's
you
know
being
actively
contributed
to
so
this
started
in
2019
and
2020
2021
all
those
years.
We
had
sort
of
different
iterations
of
this,
and
this
went
live.
B
I
think
in
21,
and
then
you
know
we
we're
this
year,
we're
in
so
it
was
like
a
pre-trained
model,
a
pre-trained
machine
learning
model
or
deep
learning
model
for
analyzing
cell,
like
doing
cell
segmentation
and
doing
some
prediction
of
movement,
and
things
like
that
now
this
year,
we're
adding
in
graph
neural
network
embeddings
and
some
other
improvements
to
the
to
the
the
methods
and
techniques
that
are
be
that
are
used
in
the
core
of
it.
B
So
you
know
a
lot
of
the
segmentation
tools
and
you
know
things
like
that:
they're
they
always
they
can
always
use
updating
and
you
know
making
them
better.
So
we
have
a
number
of
issues
in
diva
learn.
This
is,
you
know,
a
couple
things
that
are
really
kind
of
our
maintainer
who's.
I
think
it's,
my
oak
deb
and
his
brother,
my
knockdab
work
on
this
software.
B
They
were
g-shock
students,
so
we
have
a
number
of
different
issues
that
are
open
here,
that
you
know
involve
a
lot
of
technical
aspects
of
computer
vision
and
doing
things
like
unit
testing
and
other
things.
So
there's
small
things
like
that
that
that
you
know
require,
like
someone
to
just
say
I'll,
take
this
issue
and
then
work
on
a
solution
and
then
mayak
would
approve
it
and
push
it
back
to
the
repository
because
he's
the
maintainer,
but
we've
had
a
lot
of
yeah.
B
We've
had
a
lot
of
pull
requests
recently
on
this,
and
this
is
from
the
gnns
project.
So
this
is
a
couple
of
different
wataru
has
made
a
couple
of
pull
requests
here,
kim.
D
B
Has
made
a
pull
request?
Some
nets.
B
So
I've
said
hello
to
him.
He's
also
made
a
recent
pull
request,
so
we've
had
a
lot
of
pull
requests
recently
on
this
and
it's
nice
to
see,
because
you
know
it's
like
one
of
the
things
about
open
source
projects
is
that
sometimes
they
lack
sustainability
and
they
just
kind
of
fall
by
the
wayside,
but
we're
trying
to
keep
this
fresh
and
trying
to
keep
it
active.
So
this
is
definitely
something
you
know.
B
If
you
want
to
contribute
to,
is
you
know,
possibility
it's
an
active
thing
and
there's
a
lot
of
documentation
in
here
for
diva
learn.
We
also
have
a
paper,
which
is
something
that
we
keep
adding
to
as
we
make
progress.
This
was
put
out.
I
think
last
updated
at
the
beginning
of
this
year.
So
it's
like
a
summary
of
the
divo
learn
platform.
B
So
this
is
you
know.
This
is
divo,
learned,
machine
learning,
models
and
education
that
enable
computational
developmental
biology
and
it
kind
of
walks
through
the
process
of.
What's
here
what
you
can
do
with
this,
so
we
right
now
we
have
like
an
interface
that
we
can
and
we
just
had
a
we
just
had
another.
I
can't
remember
what
it
was,
but
we
had
an
improvement
on
this
as
well,
so
interacting
with
the
program
has
been
improved
upon
recently
as
well.
B
But
you
know
the
idea
is
to
take
like
a
bunch
of
images,
put
it
into
this
program
and
then
getting
a
csv
file,
which
is
a
very
simple
output
of
comma
delimited
data
that
you
can
work
with
in
your
program
of
choice,
and
then
it
gives
you,
like.
You
know,
positions
of
cells.
B
It
gives
you
boundaries
and
all
these
different
things
that
you
can
you
know
we
ideally
would
be
like
specifying
a
huge
number
of
parameters,
but
I
think
they're
only
a
couple
of
parameters
that
are
the
default,
but
we
have
like
a
nice
set
of
tools
underlying
this.
So
the
idea
is
that
you
know
it's
pre-trained
for
c
elegans.
So
if
someone
wants
to
analyze
the
c
elegans
embryo,
they
can
plop
this
in
and
they
don't
have
to
train
the
model.
B
It
just
kind
of
gives
them
a
good
answer,
but
that
doesn't
necessarily
work
for
other
types
of
embryos.
So
it's
it's.
It's
a
sort
of
a
nice
feature
and
a
drawback
of
like
pre-trained
models,
but
in
any
case,
what
we're
going
to
do
to
this
is
add
in
this
gnns
component,
so
that
the
graph
neural
network
embeddings
would
be
added
to
this.
In
addition
to
some
of
the
improvements
that
have
been
made
and
then
we'll
be
updating
this
documentation,
this
paper,
probably
by
the
end
of
this
year.
B
E
Bradley
I
there's
a
journal
that
is
interested
in
education
and
developmental
biology.
I
put
it
down.
There's
a
journal
out
of
spain.
B
All
right
so
yeah,
so
then
we
have
this
other
thing
that
we
talked
about
a
couple
weeks
ago.
This
is
for
this
symmetries
paper.
I've
been
working
on
a
abstract
for
this
and
I'm
not
quite
there.
Yet
I
don't
really
want
to
get
into
it,
but
I'm
trying
to
merge
some
of
the
stuff
we've
been
talking
about
with
embodied
cognition
in
in
in
development
and
then
also
with
development.
B
Title,
I
don't
have
one
yet
it's
just.
A
E
B
Yeah
I
was
trying
to
work
on
it
over
when
I
was
traveling
in
it
yeah
I
mean
I
got
somewhere,
but
not
where
I
wanted
to
be,
but
that's
I
just
wanted
to
keep
that
on
the
front
burner
just
to
let
everyone
know
that
that's
still
going,
and
I
got
a
interest
from
dick
and
jesse
on
this.
So
you
know
we'll
I'll
put
this
out
in
the
you
know
I'll
kind
of
send
this
out
for
review
soon
when
I
get
it
in
shape.
That
is
reasonable.
A
B
Yeah,
that's
now
I
want
to
talk
about
a
couple
weeks
ago
I
went
to
this
conference
netsi
and
we've
talked
about
networks,
a
lot
in
the
group
here
and
their
relevance
to
developmental
biology,
but
maybe
biology
more
generally,
and
I
just
wanted
to
go
over
like
what
was
there.
Maybe
what
I
learned
about
the
field,
the
state
of
the
field.
So
netsi
is
the
sort
of
preeminent
conference
in
network
science,
which
is
where
people
build
networks.
It's
not
graph
theory,
it's
it's
it's
it.
B
It's
kind
of
growing
out
of
graph
theory
and
it's
you
know
it's
characterizing
the
world
using
network
analysis.
So
this
is
where
you
have
things
in
the
world
that
are
maybe
correlated
and
you
want
to
draw
arcs
between
those
nodes.
You
characterize
something
as
a
node
and
then
you
correlate
those
nodes
in
some
way
and
you've
drawn
arc
between
the
significant
correlations
and
you
have
to
define
what
those
are.
B
And
then
you
build
this
network,
which
I
think,
if
you've
seen
like
hairballs
in
hairball
science,
or
you
know
different
things
where
you
have
like
if
you've
heard
of
small
world
networks
or
scale
free
networks
or
any
of
that,
that's
exactly
what
that
is,
and
so
the
you
know,
the
field
has
grown
quite
a
bit
in
the
last
20
years
and
they
have
they've
had
a
lot
of
innovation
in
the
field.
And
it's
and
this
year
was
supposed
to
be
held
in
china
and
then
coveted.
B
You
know
covet
is
still
a
thing
and
you
know,
and
so
they
couldn't
do
it
in
china
and
then
they
had
they
had
to
do
it
online,
which
was
good
because
you
got
a
lot
of
people,
maybe
couldn't
go
otherwise
if
they
have
that,
you
know,
and
they
have
a
nice
platform
for
people
to
interact
with.
So
let
me
get
into
the
netsi
overview,
so
this
is
netside
2022
in
shanghai.
It
wasn't
in
shanghai,
but
it
was,
I
mean
it
was
based
out
of
shanghai.
B
The
organizers
were
chinese
and
they
had
a
lot
of
chinese
groups
which
was
nice
to
see
what
they're
doing
in
the
realm
of
network
science,
and
then
you
have,
it
was
hosted
on
a
platform
called
huva,
which
is
a
nice
platform
for
doing
virtual
conferences.
B
It
you
know
it's
very
accessible,
they
basically
had
the
zoom
sessions
and
then
they
posted
them
as
recordings
to
hoova,
and
it
was
very
seamless
to
catch
up
on
the
things
you
missed,
because
the
schedule
of
course,
is
not
always.
You
know.
If
you're
in
one
place
in
the
world,
it
may
or
may
not
be
easy
to
to
make
that
time.
So
you
know,
I
think
it
started
for
me
at
like
four
in
the
morning,
so
I'm
not
going
to
get
up
at
four
in
the
morning
to
watch
it.
B
If
I
don't
have
to
it's
nice
to
go
and
see
those
talks
later
on,
but
this
is
this
is
netsi
and
they've
had
you
know
they
have
conferences
every
year
I
went
to
the
one
when
it
was
in
indianapolis
in
the
us,
and
that
was
a
nice
conference
hosted
by
indiana
university.
B
Has
a
nice
network
science
group
there
and
last
year
I
attended
and
we
did
a
talk
on
embryo
networks
there
as
well,
so
they
have
first
thing:
they
do
in
this
net
that
sci
conference
and
they
do
this
all
every
year.
Is
they
have
these
satellite
sessions
so
the
satellite
sessions?
B
B
There's
there
was
network
neuroscience,
which
was
a
very
it's,
always
a
very
big
deal
at
this
conference
because
it's
like
the
sort
of
the.
If
you
do
network
neuroscience,
that's
your
conference,
but
it's
a
satellite
of
this
conference,
so
they
they
did
that,
and
there
was
some
very
interesting
talks
there
on
brain
imaging
and
brain
imaging
networks
and
other
types
of
brain
networks.
B
They
they
wanted
to
diversify
their
presentations
because
a
lot
of
times
people
will
do
like
you
know,
fmri
networks
and
then
that's
what
a
lot
of
people
think
of
as
network
neuroscience,
but
there's
actually
a
lot
more
there,
and
you
know
there
are
other
techniques
that
people
can
use
using
other
types
of
methods.
B
So
it's
an
interesting
there's
an
interesting
session.
They
had
the
science
of
innovation,
which
is
where
you
look
at
innovation.
You
look
at
like
study,
patents,
study
like
the
innovation
of
technologies,
and
you
know
you
can
look
at
like
how
innovation
has
evolved
and
how
it's
connected
between
different
areas.
B
So
you
know
what
are
innovations
and
say,
like
automotives,
automotive
industry
related
things.
How
is
that
related
to
like
computers
and
software?
And
so
you
can
see
those
connections
between
them
using
network
theory?
You
can
see
how
they're
kind
of
related
you
know
there
there's
there
are
similarities,
there's
connectedness
there,
and
so
that's
a
that's
an
interesting
year.
B
It's
actually
a
very
old
area,
because
you
know
there
was
a
lot
of
work
being
done
in
like
the
30s
and
40s
on
this,
and
so
you
know
that
that
work
has
kind
of
continued
on
forward
to
the
modern
day.
B
Then
there
was
women
in
narrow
network
science
and
diversify
network
science,
and
that's
actually
also
a
very
well
organized
satellite
and
it
happens
every
year
and
they
had
it
was
basically,
you
know
women
in
network
science,
and
so
it
was
very
diverse
in
terms
of
the
talks
that
were
happening
and
that
was
another
satellite.
B
Then
there
was
net
biomed
which
I
attended
again.
This
is
more
about
like,
like
we
talked
about
some
of
the
networks
in
systems
biology
and
in
computational
biology.
So
there
was
a
lot
of
like
you
know,
protein
protein
interaction
networks
and
other
types
of
networks
that
are
relevant
to
biomedicine
and
systems
of
biology.
B
So
they
were
doing
a
lot
of
interesting
work
there.
Then
there
was
a
netsi
education,
so
this
is
like
education
in
network
science.
Different
people,
you
know,
talked
about
their
initiatives
for
educating
people
on
this
topic,
multi-layer
networks,
which
are
networks
where
you
have
multiple
layers.
B
So
this
is
useful
if
you're
thinking
about
like
mapping
the
genotype
to
the
phenotype
or
if
you're
mapping,
you
know
different
layers
in
time.
So
if
we
have
like
a
network
in
a
you
know
at
time,
zero
versus
a
network
at
time,
two
or
three
we
can
model
each
time
step
as
a
network
of
its
own
and
they
you
know
they
have.
You
have
this
multi-layered
network.
B
Where
there's
ostensibly
connections
between
the
layers,
then
there
are
higher
order,
topologies
and
dynamics,
and
I
think
I've
talked
to
gia
hong
about
this,
how
you
can
incorporate
topological
data
analysis
and
topological
approaches
into
networks
in
the
connected
networks,
and
then
you
know
that
automatically
we
do
to
dynamics
because
you're
also
interested
in
temporal
aspects
of
it,
so
that
I
didn't
go
to
this
session
the
satellite
session.
B
But
it
was
a
pretty
interesting
stuff.
I
looked
over
the
you
know
the
list
of
talks,
and
then
there
was
this
paper
unwind
and
that's
how
a
paper
comes
about,
and
I
thought
that
was
a
nice
satellite
to
have
for
conference,
because
it's
basically
telling
people
how
to
write
a
paper,
how
you
can
conceptualize
a
paper
from
like
a
very
fuzzy
concept
to
like
a
published
paper
and
they
had
some
journal
editors
in
attendance.
B
So
people
could
ask
questions
about
that
process,
but
that
was
a
nice
way,
nice
session
to
have
and
probably
pretty
useful
for
young
investigators.
So
this
is
one
of
the
things
I
I
presented
at
the
network
site
at
the
net
biomed
session.
B
B
So
you
know,
if
you
have
things
that
don't
happen
very
often
or
if
you
have
things
that
are
sort
of
a
diffuse
sort
of
measurement.
B
So
I
have
this.
It's
like
a.
I
don't
know
how
many
minutes
it's
not
that
long,
but
I
have
a
video
of
this
presentation
on
the
youtube
channel
for
diva
worm.
B
The
second
one
is
the
circuitous
connecto
modeling
for
developmental,
developing
sensory
motor
complexity,
and
this
is
another.
This
is
a
poster
I
presented
at
network
neuroscience.
B
This
is
some
some
of
the
stuff
that
we've
done
with
agents
and
and
growing
connectomes,
but
also
some
other
work
from
a
long
time
ago,
again
on
some
things
related
to
how
to
build
up
a
network
that
isn't
like
optimal
but
actually
sub-optimal.
B
So
the
idea
is,
is
that
you
have
these
sub-optimal
structures
in
nature.
You
know
in
different
organ
systems
that
aren't
like
optimized
and
they're
deliberately
sub-optimal,
and
so
that
it
took
that
as
an
inspiration
for
this
sort
of
modeling
project.
Where
built
these
circuitous
circuits,
which
is
to
say
that
they're
not
optimal,
they're,
actually
sub-optimal
the
rules
enforce
things
to
be
like
for,
say,
a
signal
to
take
the
longest
group
possible
through
a
network
or
to
build
a
network
as
complex
as
possible
and
still
work.
And
so
the
idea
is.
B
And
so
the
main
talks
were
six
six
talks
in
each
of
these
categories
of
15
minutes
each.
Some
of
these
you
know
had
like
multiple
sessions
across
days,
so
the
theory
and
structure
session
was
six
talks
on
one
day
and
then
there
was
a
theory
in
structure.
Two
on
another
day
and
so
forth,
so
theory
and
structure
is
very
important
in
networks,
and
this
is
really,
as
it
says,
theory
which
is
like
you
know.
B
How
do
we
interpret
some
of
the
things
that
we
see
in
networks,
but
also,
how
do
we
build
networks?
What's
the
theory
behind
some
of
the
methods
that
are
used
and
the
structure,
so
in
networks,
you
have
structures
like
communities,
modules,
things
like
that,
and
so
a
lot
of
the
talks
focused
on
that
these
are
you
know
across
fields
like
these
use
data
from
different
areas?
It's
not
just
like
one
area
of
science.
It's
multiple
areas
could
be
biology.
It
could
be
economics,
it
could
be.
You
know,
power
the
power
grid.
B
There
are
all
sorts
of
places
you
can
get
data,
but
you
just
kind
of
unify
a
lot
of
that
with
this
common
set
of
tools
and
theories,
it's
very
it's
very
much
a
science
of
complexity,
in
that
it's
it's
agnostic
to
the
field.
It's
just
kind
of
like
this
is
how
you
build
these
networks
and
they
should
work
across
fields.
B
Then
there
was
social
medias
of
people
who
analyze
social
media
interactions.
This
is
one
of
the
things
they
often
use
the
twitter
api
for
this
to
get
like
data
large
data
sets
and
then
look
at
how
people
are
discussing
topics
or
interacting.
B
So
it's
social
networks,
the
science
of
innovation.
Again
this
is
where
you're
looking
at
innovation
in
different
fields
and
you're,
looking
at
how
it
connects
there
are
different
dimensions
of
networks
of
temporal
and
spatial,
so
treating
networks
is
specifically
spatial
things
that
you
know
they're
nodes
in
one
place
and
space
versus
another,
because
a
lot
of
network
science
is
really
kind
of
divorced
from
space.
It's
just
kind
of
like
here's,
a
network,
there's
a
pretty
graph,
it's
sort
of
a
distance
metric,
but
we
don't
really
know
where
it
is
in
space.
B
It
doesn't
matter
and
these
kind
of
networks
it
does
matter
very
much
where
things
are
in
space
as
well
as
time.
So
there
are
dimensions
to
networks
that
you
can.
People
are
developing
techniques
for
this.
There
is
some
work
on
spatial
analysis
and
embeddings,
which
is
again
where
you're
embedding
the
data
into
a
certain
geometry.
B
You
have
field
specific
talks
on
economics
and
biology.
I
think
those
were
the
two
specialized
areas.
My
talk
was
in
the
biology
session,
but
you
know
this
is
where
you
take
this
these
networks
and
they
apply
to
all
fields
but
you're
using
it
in
a
specific
field.
So
it
was
nice
to
have
that
those
field
specific
sessions
and
then,
finally,
they
had
a
statistics
and
ai
session,
which
was
involved
a
lot
of
machine
learning
and
inference
techniques.
B
So
one
of
the
things
that,
like
I
found
about
this
conference,
is
that,
unlike
past
years,
they
really
made
a
point
to
incorporate
a
lot
of
statistical
methods,
a
lot
of
other
statistical
methods
from
other
areas,
machine
learning
methods
and
the
like.
B
So
my
talk
was
main
talk
session
talk,
biological
networks,
too,
was
hyper.
Graphs
demonstrate
anastomoses
during
divergent
integration,
and
this
was
a
talk
that
I
I
don't
know
if
I've,
I
think
I
showed
the
slides
in
an
earlier
session,
but
all
of
these
all
these
sli,
all
these
talks
and
posters
that
I've
posted
here
are
on.
Actually,
I
think
this
one
is
on
the
orthogonal.
Research
and
education
lab
youtube.
This
one
is
on
the
diva
warm
youtube,
so
you
can
get
a
sense
of
like
what
was
going
on
in
that
talk.
B
It
was
an
interesting
talk
because
it
took
embryogenesis
and
it
broke
it
down
into
this
idea
of
hypergraphs,
which
are
where
you
have
these
nodes.
But
the
nodes
contain
a
lot
of
different
things,
a
lot
of
different
things
of
the
same
category
inside
them,
and
so
then
you
treat
the
graph.
The
the
network
is
sort
of
a
graph
that
has
multiple
parts
within
a
node,
multiple
arcs
within
an
arc,
and
you
you
can
do
a
lot
of
different
interesting
analyses
with
this.
So
you
have
like
almost
like
this.
B
This
scale
you
can
have
things
at
different
scales.
You
can
look
at
like
a
group
of
cells,
doing
something
versus
single
cells,
doing
something
and
those
are
what
we
call
hypergraphs
and
so
there's,
I
think,
there's
a
lot
of
potential
for
hypergraphs
and
maybe
we'll
talk
about
that
in
another
meeting.
B
B
There's
a
lot
of
like
you
know,
a
lot
of
the
new
techniques
in
data
science
and
machine
learning
and
deep
learning
are
kind
of,
like
you
know,
been
developing
for
a
while
and
they're
kind
of
going
into
this
area,
because
you
have
people
maybe
working
between
the
areas
where
there
are
a
lot.
There's
a
lot
of
computational
potential
here
between
say
like
deep
learning
and
network
science,
and
we
see
that
with
graph
neural
networks.
Just
as
an
example,
there
is
also
a
focus
on
topological
data
analysis
and
the
light
techniques.
B
So
they
there
were
a
lot
of
talks
about
things
like
taking
what
they
call
a
motif,
which
is
like
a
a
triangular
shape
in
the
network,
with
three
nodes
connected
into
a
triangle
and
treating
those
is
what
they
call
simplices,
which
are
like
treating
that
as
an
actual
triangle
and
then
taking
the
shapes
and,
like
you
know,
thinking
about
them
is
tilings
or
things
like
that.
B
So
you're
taking
the
network
apart
and
you're,
finding
shapes
within
the
network
that
are
these
simplicities
and
the
idea
is
that
the
way
they're
connected
together
like
that
that
can
tell
you
something
about
what's
going
on,
then
there
are
connections
between
heredity,
hyper
graphs
and
other
techniques,
so
this
is
just
saying
that
there's
always
room
to
incorporate
techniques
from
other
fields.
A
Things
but
also
a
lot
of
algorithmic
work
and
well
javascript
for
media.
Maybe
a
lot
of
js.
A
B
B
So
great,
what
do
you?
What's
your?
What
are
your
interests?
I
mean:
what
do
you
do
besides
javascript?
How
do
you
apply
it
to
things.
A
Oh
god,
I
mean
it's
what
I'm
a
I'm
a
renaissance
man
right
now,
I'm
into
pretty
much
everything,
but
now
so
right.
Now,
I'm
building
I'm
building
like
wearable
tech,
so
for
for,
like
narrow
tech
and
then
for
rehab,
and
then
and
now
we're
just
trying
to
use
web
to
build
all
these
open
source
tools
so
that
it's
really
easy
for
us
to.
You
know
develop
with
like
a
community.
A
So
then
like
right
now,
I'm
working
on
like
a
whole
visual
programming
interface
and
like
like
a
like
a
flowgraph
programming
framework
and
all
this
stuff
in
javascript,
and
then
we're
writing
yeah
like
rehab
software.
With
this
and
some
educational
tools
for
like
high
schoolers,
I
don't
know
my
interests
are
everywhere,
though,
but
I'm
trying
to
learn
all
this
crazy
math
over
time.
So
we.
B
A
Get
into
you
know
like
some
of
this,
that
you
guys
do
with
the
devil.
That's
not
you
know,
don't
let
josh's
films
off
short,
but
also
on
the
g32
and
128
as
well.
I
mean
so
like
these.
Are
you
know,
nice
schematic
hardware,
designs
that
that
we're
yeah
we're
looking
to
use
for
psychiatric
applications
like
like
addiction,
and
you
know
other
other
centers
yeah.
A
B
B
Oh,
that's
good,
yeah,
well,
yeah
you're
welcome
to
join
our
meetings
anytime
and
well.
You
know
I
mean
if
you
want
to
collaborate,
we
can
talk
about
some
things.
That
way,
I
just
told
a
lot
the
best
way
to
to
learn
what
we're
doing
is
to
attend
the
meetings
and
kind
of
ask
questions
about.
What's
going
on,
so
that's
you
know
so.
C
Yeah
I
just
just
joined
today
as
well,
so
just
like
yeah,
an
opening
from
bradley
like
yeah
earlier
before
yeah
nice
to
meet
you.
B
Yeah,
so
I
think
that's
it
for
today
we
had
anything
else.
We
wanted
to
talk
about
next
week,
we'll
be
giving
more
updates
on
gsoc
and
coming
closer
to
the
deadline
on
that.
So
I
guess
my
advice
to
gsoc
students
is,
you
know,
work
on
the
documentation,
as
well
as
the
finishing
projects,
and
just
don't
worry
about
getting
it
perfect,
just
get
it
to
a
state
where
we
can.
B
You
know
where,
if
you
submit
your
project,
it
can
be
run
because
that's
one
of
the
things
they
do
in
gsoc
is
when
you
submit
the
project
it
has
to
be,
they
have
to
be
able
to
run
it
to
get
it.
You
know
due
to
pass,
so
it's
just
the
matter
of
making
sure
it
runs
and
it
doesn't
have
to
be
perfect.
We
can
work
on
that
after
the
project's
end,
like
especially
when
we
you
know
we're
going
to
incorporate
this
into
divalern.
B
We
want
to
be
able
to
we're
going
to
probably
have
to
make
some
changes
to
it
or
put
in
you
know
some
things,
but
that's
that's
that
doesn't
need
to
be
done
for
the
gsoc
project
and
you
know
again.
B
A
A
He's
a
big
eeg
algorithm
guy,
okay,
yeah
yeah
yeah.
He
runs
the
cuban
neuroscience
center.
A
Also
it's
in
in
shang
view
I
forget
the
name
of
the
university,
but
but
runs
the
center
there
as
well,
but
I
I
was
just
gonna:
ask
them
the.
B
I
don't
know
of
any
labs
that
are
using
it.
I
mean,
I
know
they're
like
there
are
a
lot
of
you
know
several
hundred
downloads
of
the
software,
but
it's
you
know
I
don't.
I
don't
know
who's
actually
using
that's
one
of
the
things
I
I
am
always
like,
like
a
nice
surprise
when
you
find
out
who's
using
it,
but
there's
no
good
way
to
really
get
a
census
on
that.
A
You
know
that's
why,
like
you
know,
some
groups,
like
I
mean
again
going
from
managing
like
spm
and
and
eeg
lab,
make
you
make
you
go
through,
not
a
paywall,
but
just
a
like
give
us
your
name
and
an
email
address,
so
that
we
can.
We
can
collect
those
for
when
we,
when
we
submit
grants
and
things
yeah.
A
Yeah,
but
I
was
just
wondering
you
know
it
would
be
nice
to
reach
out
to
some
some
steel
against
labs
and
you
know
see
like
yeah.
Could
they
you
know,
do
they
know
about
it
and,
and
is
it
would
it
fix
any
problems
that
they
you
know?
Would
it
replace
anything
that
they
they're
currently
using?
Would
it
be
better
than
again
just
curious
of
you
know
how
much
it's
getting
used
by
by
those
groups.
A
B
B
B
So
this
is
the
title.
This
is
an
e-wife,
it's
by
a
bunch
of
people
at
the
department
of
biochemistry
and
biophysics
at
ucsf,
and
so
wallace
marshall
was
a
big
name
here
and
one
one
of
the
big
names
in
this
paper,
and
so
the
title
is
modular
cascade
like
transcriptional
program
of
regeneration
and
stentor.
So
they
took
this
organism
they're.
B
Looking
at
the
transcriptional
program,
it's
very
simple
organism
and
they're
going
to
describe
the
regeneration
program
as
it
as
they're,
referring
to
it
and
what
that
looks
like
so
the
abstract
reads:
the
giant,
solute
stentor
corialis
is
a
classical
model
system
for
studying,
regeneration
and
morphogenesis
in
a
single
cell.
So
we've
talked
about
single
cell
models
before
this
is
a
ciliate,
so
this
is
different
than
like
say
some
of
the
other
models
we've
looked
at
like
in
algae
and
other
things.
B
The
interior
of
the
cell
is
marked
by
an
array
of
cilia
known
as
the
oral
apparatus
which
can
be
induced
to
shed
and
regenerate
in
a
series
of
reproducible,
morphological
steps,
so
they're
regenerating
the
oral
apparatus
in
the
cell,
and
this
is
the
regeneration
aspect
of
it.
So
this
is
like
regeneration
and
or
at
least
that's
a
nice
analogy
to
it.
It's
regenerating
a
body
part
which
is
actually
a
part
of
the
cell,
but
in
any
case
it's
it's
a
reproducible
thing.
B
So
it's
some
sort
of
program,
that's
generating
regenerating
this
part
of
the
cell,
so
the
series
of
reproducible,
morphological
steps
has
actually
been
shown
to
require
transcription.
So
we
don't
necessarily
well.
We
know
that
transcription
is
involved
in
a
lot
of
different
types
of
regeneration,
but
this
is
they're
kind
of
tying.
This
to
specific
transcriptional
thing.
You
know
mechanisms
if
a
cell
is
cut
in
half
each
half
regenerates
in
intact
cells,
so
this
is
reminiscent
of
some
of
the
things
of
flatworms.
B
Where,
if
you
take
this
worm,
you
cut
it
in
half,
it
generates
two
new
worms.
If
you
gener,
if
you
take
a
single
cell
and
you
replate
it,
you
can
get
an
entire
organism,
but
we
know
that
in
a
flat
worm.
This
is
because
the
cells
are
tote,
what
they
call
totipotent,
which
means
that
they
can
generate
into
all
the
constituent
cells
of
an
adult
flatworm,
but
it
also
suggests
that
there's
some
sort
of
organizing
principle
that
allows
it
to
form
this
body.
B
B
B
So
now
there's
steps
of
in
the
regeneration
process.
The
oral
apparatus
is
regenerated
and
there's
a
recovery
after
wounding,
so
they're
different
steps
just
like.
If
you
cut
your
arm,
you
have
different
steps
of
wound
healing
and
that's,
of
course,
regeneration
of
the
skin
cells
and
the
muscle
cells
and
other
things
in
that
region
of
where
you
cut
your
arm
by
measuring
gene
expression.
After
blocking
translation,
we
show
that
the
sequential
waves
of
gene
expression
involve
cascade
mechanism,
which
later
waves
of
expression
are
triggered
by
translational
translation
products
of
early
express
genes.
B
So
this
happens
in
a
lot
of
transcriptional
systems.
You
have
what
they
call
early
expressed
or
even
early
immediate
genes
that
respond
when
there's
a
stimulus.
So
if
there's
a
stimulus
like
you,
you
get
your.
You
cut
your
arm,
their
genes
that
get
upregulated
they
get
expressed
at
a
greater
level
when
you
know
they're,
they're
ones,
that
just
kind
of
respond
to
that
stimulus
and
they
do
maintain
things
involving
maintenance
and
setting
up
other
cascades
that
come
later
and
then
you
have
later
on.
B
You
have
genes
that
are
actually
involved
in
this
process
of
repair
and
of
things
like
that.
So
there's
this
time
series
of
gene
expression,
it's
not
just
that
you
get
this
massive
upregulation
of
the
genes
or
that
this
this
circuit
turns
on.
You
know
just
automatically
and
then
fixes
things,
there's
a
temporal
aspect
to
it,
and
so
that's
what
they're?
B
E2F
is
involved
in
the
completion
of
regeneration,
but
is
dispensable
for
earlier
steps.
This
work
allows
us
to
classify
regeneration
genes
into
groups
based
on
their
potential
role,
for
regeneration
and
distinct
cell
regeneration
paradigms
and
provides
insight
into
how
a
single
cell
and
coordinate
complex,
morphogenetic
pathways
to
regenerate
missing
structures.
B
So
you
can
see
that
we
don't
talk
about
regeneration
as
much
in
this
group
as
we
talk
about
things
like
morphogenesis
and
embryogenesis,
but
you
can
see
that
there
are
a
lot
of
similarities
in
the
process
that
you
need
to
basically
regenerate
something
that's
been
lost.
It
has
to
match.
What's
there
now,
so
this
is
kind
of
like
maybe
local
morphogenesis
developmentally.
B
There
are
different
processes,
but
they
have
a
lot
of.
They
have
a
lot
of
parallels.
Interestingly,
though,
this
is
an
interesting
thing
to
see
in
a
single
cell
organism,
because
we
usually
associate
development
with
multicell
organisms.
Sometimes
single
cell
organisms
have
a
developmental
stage,
but
it's
it's
kind
of
a
you
know.
B
How
do
you
get
from
like
dna
to
a
you
know
an
asymmetrical
phenotype,
so
this
is
an
interesting
thing
to
see
here
in
the
literature,
because
it
really
kind
of
nails
down
what
the
program
looks
like,
but
it
also
shows
in
a
single
cell
organism.
B
So
while
much
is
known
about
the
molecular
composition
of
cells,
the
mechanism
by
which
these
components
are
arranged
in
to
complex
patterns
and
structures
is
far
less
understood.
So
how
does
a
cell
create
and
maintain
pattern?
So
this
is
something
that
you
know.
People
have
looked
at
for
a
long
time
and
people
have
been
interested
in
regeneration
and
in
humans,
but
also
we
have
animal
models
for
this,
but
so
it's
are
a
new
model
for
this,
because
it's
a
very
simple
organism,
but
it
still
has
a
lot
of
these
mechanisms.
B
So
if
we
look
at
some
of
these
images
here,
we
see
that
this
is
a
stentor.
We
have
these.
This
is
the
oral
apparatus
at
the
top,
the
anterior
end
and
the
hold
fast
at
the
posterior
end,
and
it
kind
of
looks
like
this
thing:
that's
this
cone
this
conical
shape
and
you
can
see
that
then
it's
been
wounded
and
then
it
needs
to
regenerate
its
oral
apparatus.
B
So
you
can
see
that
there's
this
stage
of
of
different
things
and
it
kind
of
looks
like
some
organ
is
being
regenerated,
but
it's
actually
just
part
of
a
single
cell.
You
can
actually
cut
it
in
half
and
the
posterior
end
of
the
top.
The
anterior
end
will
regenerate
and
then
the
posterior
part
will
regenerate
an
anterior
part.
B
We
have
these
different
categories,
so
we
have
genes
involved
in
tail
regeneration,
bisection
response,
anterior
regeneration,
general
regeneration
and
then
sucrose
shock.
So
these
are
all
expressed
during
this,
this
sort
of
environmental
challenge
or
this
this
physiological
challenge-
I
guess
in
this
case,
so
it's
being
cut
in
half
or
it's
being
wounded
in
some
way,
and
there
are
these
genes
that
are
responding
that
are
upregulated,
and
this
is
what
their
categories
are.
The
functional
categories.
B
B
They
were
able
to
show
that
there's
this
time
course
to
the
regeneration
that
it
happens,
that
you
know
different
genes
are
involved,
so
they
actually
defined
six
sets
of
differentially
expressed
genes
genes
expressed
in
both
sucrose
shock
and
regenerating
posterior
halves,
which
we
take
to
indicate
genes
required
for
oa
regeneration,
sucrose
shock,
specific,
which
we
interpret
as
reflecting
aspects
of
oa
regeneration
in
sucralose
shocked,
but
not
bisected
cells.
So
they
did
a
suprashock
treatment
versus
just
bisecting
the
cells,
so
they
were
able
to.
B
So
this
is,
you
know,
there's
an
osmotic
response
that
involves
these
suprashock
genes,
responding
to
the
suprashock
genes,
expressed
only
in
regenerating
posterior
halves,
which
we
interpret
as
relating
to
regeneration
of
anterior
structures
other
than
the
oa
genes,
express
only
regenerating
anterior
halves,
so
they're
genes
that
are
involved
in
both
the
posterior
and
anterior
specific
regeneration
and
then
genes
expressed
in
both
regenerating
half
cells,
but
not
sucrose
shock.
So
these
are
bisection,
specific
and
then
finally,
genes
expressed
in
all
three
samples
which
we
take
to
indicate
general
regeneration
genes.
B
Some
of
the
genes
are
shared
between
different
categories
and
some
are
not
so
generally
the
way
we
look
at
this
is
we
look
at
this
as
differentially
expressed
genes.
We
look
at
like
some
sort
of
housekeeping
gene
or
standard,
and
then
we
look
at
the
thing,
that's
being
upregulated
or
down
regulated,
and
we
can
make
a
judgment
as
to
whether
that's
a
significant
deviation
or
a
significant
amount
of
expression,
so
these
genes
that
they
identified
there
was
some
sort
of
test
done
where
they
had
a
cut
off.
B
For
you
know,
expression,
elevated
or
decreased
expression
relative
to
some
standard,
and
so
this
is
how
they
defined
these
differentially
expressed
genes
and
are
different
equations
for
doing
this.
Basically,
you
take
the
level
of
expression.
You
normalize
it
by
the
control,
which
is
usually
some
sort
of
housekeeping
gene.
That's
not
involved
necessarily
in
this
whole
process,
and
then
you
make
there's
some
fold
measurement.
So
it's
like
two
fold
up
regulation
or
two
full
differentially
expressed
regulation
is,
like
you
know,
a
part
of
that
fraction.
B
So
you'll
get
a
you
know,
you'll
get
a
number
that
you
can
attach
to
it.
The
question
is:
what's
significant,
is
it
one
fold?
Is
it
twofold
et
cetera?
So
this
is
the
thing
that
we
wanna
we're
trying
to
do
and
then
once
we
get
this
subset
of
things
that
are
significant,
then
we
can
cluster
them
into
these
functional
categories.
B
So
they
go
through
a
lot
of
these
different
categories.
They
have
these
heat
maps,
which
show
the
differential
expression
over
time
and
across
genes,
and
they
have
these
clusters
that
they
generate
so
they're
doing
this
over
a
time
series
as
well.
So
they
have
like
these
different
time
points
of
measurement
and
they
can
show,
as
I
showed
in
the
preview
one
of
the
previous
panels,
that
there's
this
process
that
unfolds
over
time
and
that
if
you
have
an
anterior
versus
posterior
regeneration
potential,
that
those
are
different
in
terms
of
time.
B
So
this
is
a
lot
of
data
in
this
paper,
a
very
nice
piece
of
work
and
they
kind
of
go
through
some
of
these
other
categories
that
we
just
mentioned,
and
so
don't
know.
B
B
Okay
and
then
the
discussion
talks
about
modularity
in
the
regeneration
program.
So
one
question
we
ask
in
molecular
biology
is
what
are
the
modules
underlying
different
processes,
so
we
have
different
modules
in
this
program
right.
You
don't
have
just
the
program
that
activates,
you
have
different
modules
that
occur
so
there's
a
collection
of
distinct
part,
specific
processes
and
the
thing
about
modularity
is
that
if
things
get
knocked
out,
sometimes
that
you
know
limits
the
effectiveness
of
that
program.
B
But
of
course,
if
there
are
overlaps
between
the
different
modules,
where
there
are
different
ways
in
which
they're
they
evolve,
so
that
they're,
you
know
critical
genes
and
less
critical
genes
and
the
critical
genes
are
protected
by
some
sort
of
purifying
selection
or
something
that
gets
rid
of
all
mutations.
B
B
So
the
fact
that
general
regeneration
module
contains
the
fewest
genes
out
of
all
the
modules
identified
suggests
that
regeneration
does
not
represent
a
single
master
program
of
expression,
but
rather
a
composite
of
distinct
expression,
modules
or
subroutines
specific
for
regenerating
individual
parts
of
the
cell.
Somehow
the
cell
must
recognize
which
parts
it
is
missing
and
trigger
the
appropriate
module
to
restore
that
part.
So
that's
an
interesting
take
on
this
that
we
can
see
from
those
categories
of
transcriptional
clusters
that
you
have
these
different
groups
and
you
can
map
those
different
groups
to
different
genetic
modules.
B
Yeah
some
cells
are,
or
some
genes
are
responsible
in.
You
know
for
more
than
one
function,
and
it
has
a
lot
of
consequence
for
things
like
robustness
and
other
and
redundancies
in
the
circuit.
And
so,
if
you
want
to
read
more
about
this
paper,
go
ahead
and
read,
there
are
a
lot
of
interesting
things
here,
comparing
it
to
other
studies,
and
so
thank
you
for
paying
attention
all
right.