►
Description
GSoC Community Period (week 2) continues with some words from Tom Portegys on bio-inspired AI and robotics. GSoC intern updates. Discussion of how past modeling efforts can be used to develop a model of regenerating networks. Discussion of neural organoids and a paper on Body Plan Identity in the evolution of development. Attendees: Susan Crawford-Young, Morgan Hough, Jesse Parent, Harikrishna Pillai, Tom Portegys, Richard Gordon, Wataru Karakami, and Bradly Alicea
A
B
Oh
and
I
got
four
four
new
baby
axolotls
that
are
six
months
old,
and
so
by
christmas
they
should
be
a
year
old.
Who
could
yes-
and
I
I
put
winnie
with
chomper
and
stickers,
but
nothing's
happening.
C
D
D
A
So
welcome
to
the
meeting
karan
said
he
might
be
able
to
join
a
little
bit
later,
so
that'll
be
good.
A
C
Oh
hi
hi.
I
I
discussed
with
jihan
how
to
about
how
to
start
the
project
and
we
decided
to
start
with.
C
Deciding
what
inputs
and
output
formats
we
we
are,
we
are
sorry
I
we
we
expect
in
each
modules.
C
So
so
we
are
going
to
start
with
check
what
outputs
the
devil
learns
and
other
other
libraries
like
people
learn,
provides.
A
All
right:
well,
I'm
glad
you
you
two
were
able
to
get
together
and
I
told
you
hung
last
week
that
we
needed
to
get
together
on
that.
Just
so
you
don't
replicate
efforts
in
that
year,
especially
with
this
project,
because
it's
something
you
know
it's
like.
A
I
know
we've
been
talking
about
the
the
spherical
map
for
a
while,
but
this
is
something
new
and
I
am
not
quite
sure
how
it's
going
to
turn
out
or
you
know
I'm
not
quit
I
mean
I
know
what
the
goal
is,
but
this
is
something
it's
going
to
be
very
trial
and
error.
So
thank
you
for
meeting
with
him
and
did
you
have
any
questions
about
any
of
the
materials
that
I
posted
in
the
slack
for
the
community
period.
C
No
for
now
I
I
don't
have
any
specific
problems
so
yeah,
I'm
still
catching
up
all
right.
A
Time,
like
the
community
period,
is
a
good
time
to
get
all
this
straightened
out,
like
all
the
things
that
you
need
to
do
to
get
started
on
the
project,
so
anyways
yeah,
if
you
have
any
other
questions,
put
them
in
the
slack
and
I'll
be
I'll
look
forward
to
talking
to
you
about
it,
so
jesse
actually
is
going
to
be
a
co-mentor
for
the
gnn's
project,
so
it'll
be
wataru
and
jiahung
who
are
going
to
be
in
that
project
and
jessie's
works
with
me
on
some
other
things,
gsoc
related
as
well
as
another,
a
bunch
of
other
projects
that
we're
working
on
and
so
he's
a
lab
manager.
A
So
he's
good
he's
good
at
sort
of
organizational
stuff
he's,
also
very
good
at
he's
done
a
lot
of
stuff
with,
like
you
know,
computational
neuroscience,
and
things
like
that,
so
he
can
try
to
help
you
with
some
things.
I
wouldn't
you
know
I
mean
really
technical
stuff.
Maybe
not!
I
don't
know
how
you
feel
about
that
jesse,
but
definitely
like
you
know.
If
there's
a
if
there's
a
question,
if
there's
some,
you
know
something
you
need
to
find
out
about.
A
You
know
you
can
get
in
touch
with
me
or
jesse
and
then
you
know
we'll
be
planning
other
things,
especially
during
the
coding
period,
to
help
you
work
on
these
projects.
So.
A
F
Yeah,
so
I
so
last
week
I
tried
to
go
forward
with
the
nerf
algorithm.
I
also
talked
with
current
so
right
now
we
are
working
on
that
like
how
do
we
use
it
exactly
for
the
android
problem
that.
A
Right,
so
how
are
you
coming
with
data?
Are
you
all
set
with
data,
or
are
you
needing
different
data
sets,
or
I
don't
know
how
that
got
resolved.
F
Right
now
I
have
like
some
of
the
pepper
datasets
and
the
pearl
dataset,
which
susan
had
said.
I'm
using
that.
B
Yeah
these
are,
I
got
these
from
all
holes,
perforating
and
they're
spherical
holes,
so
I'm
I'm
hoping
to
put
them
on
on
my
ball
bearings
or
something
and
just
put
paint
on
there
and
then
move
to
another
spot
and
put
paint
so
that
there's,
I
don't
know
some
proper
pattern
so
about
two
sizes.
B
So
I'll
play
with
this.
I
have
I'm
sent
for
some
ink
that
resin
alcohol
ink,
which
is
apparently
better
for
not
running
so
I
can.
I
can
do
that,
and
I
I
mean
I
could
also
use
larger
spheres
for
some
of
this,
so
that
I
may
be
easier
to
paint
something
on
with
a
larger
sphere.
Is
all
right?
B
B
Yeah
well
I'll,
try
all
kinds
of
things
in
here
willing
to
try
things
very.
A
Fast
yeah-
and
this
is
for
what
is
this
for
like
well.
B
Taking
pictures
with
my
ball
microscope
to
try
to
get
something
that
you
can
actually
match
so
that
you
can
image
it
okay,
because
these
these
things
are
just
very
uniformly
white
and
so
they're.
Not
it's
not.
E
Yeah
yeah,
so
if
you
trying
to
trying
to
get
texture
on
a
two
millimeter
diameter
sphere
is
rather
difficult.
A
Yeah
yeah
otaru,
I
was
wondering
if
you
and
gia
hong
had
discussed
your
data.
Sets
that
you're
going
to
use.
Are
you
do
you
have
enough
data
or
how
are
you
pairing
with
that?.
C
Yo
we
are
now
checking
the
data
sets
you
you
gave
us.
A
C
And
yeah,
we
don't
have
any
specific
details
to.
We
haven't
decided
which
data
sets
to
approach
yeah
yeah.
A
A
And
in
past
years
I
mean
people
have
used
the
you
know:
they've
used
well,
the
ones
that
I
gave.
I
I
sent
you
links
to.
Those
are
usually
sufficient
to
do
benchmarking
and
things
like
that,
so
you
know
there's
the
the
embryo
that
you
have
and
it
you
know
the
cells
divide
over
time.
So
you
have
these
movies
and
then
each
movie
has
a
frame
and
the
frame
has
like
a
certain
number
of
cells
in
it.
A
Some
of
them
are
dividing
some
of
them
are
just
cells,
and
then
we
have
information
in
c
elegans
we
have
information
on
the
identity
of
the
cells,
so
those
cells
can
be
labeled.
A
You
know
with
with
this
information,
so
this
is
something
that
you
know
is
nice,
because
a
lot
of
other
species
of
a
lot
of
other
species
or
a
lot
of
other
developmental
systems.
Don't
have
this
information,
so
you
have
labeling
pretty
much
built
into
this.
So
that's
good,
and
then
you
know
this
is
again
like
we've
done
different.
A
You
know
different
ways
of
doing
this
throughout
the
years
where
we've
worked
on,
you
know
how
to
get
the
thing
to
recognize
cells
and
how
to
set
up
a,
I
think
it
was
2017.
I
think
we
did
a
gsoc
project
where
we
did
some
work
on
this,
where
we
were
connecting
the
lineage
tree
with
the
images.
A
So
it's
it's
something
that
can
be
done
so
I
mean
you
know
there
are
different
ways
you
can
approach.
This
is
what
I'm
saying
so,
just
let
me
know
if
you
have
any
questions
about
anything
and
so
that
you
can,
you
know
you
can
move
forward
on
it.
A
Okay,
well,
thank
you
for
the
those
updates.
So
this
past
week,
like
friday,
I
gave
a
talk
at
the
google
summer
of
code
contributors
summit,
so
this
is
for
all
contributors,
mentors
and
people
running
the
pr
program.
So
there
are
a
lot
of
really
and
I
have
a
link
to
it
in
the
slack.
I
think
I
put
it
in
there
friday
and
it's
a
live.
You
should
be
able
to
access
the
live
stream.
A
A
I
was,
I
gave
a
talk
on
some
of
the
educational
initiatives
we've
been
doing
over
the
past
several
years.
So
that's
an
interesting
that
was
an
interesting
session
I
actually
met
up
with.
I
was
in
this.
I
shared
a
session
with
arnab
banerjee,
who
worked
with
the
diva
warm
group
back
in,
I
think,
2017
or
2018,
and
so
it's
kind
of
fun
to
watch
him
as
he's
moved
on
in
his
career
he's
now
working
at
credit
suisse,
which
is
a
bank
and
he
started
out
an
open
source.
A
They
wanted
his
open
source
expertise
originally
and
then
he's
moved
on
to
project
management
in
those
kind
of
roles.
So
that's
nice
to
see
him
move
on
in
his
career.
Of
course,
it
wasn't
all
due
to
gsoc,
but
I
think
that
helped
him
maybe
getting
his
foot
in
the
door.
Maybe
it's
helped
him
in
his
career.
A
He
was
definitely
at
the
contributor
summit,
so
you
know
that
they've
and
a
lot
of
people
found
that
in
a
nice,
interesting
talk-
and
there
are-
you-
know,
other
talks
on
licensing
talking
about
how
open
source
licenses
are.
You
know,
depending
on
what
you
want
to
do
with
your
software.
A
A
So
you
know
people
create
an
open
source
project
in
a
corporation
and
they
put
a
gpl
license
on
it.
It
can
pick
up
dependencies
that
are,
you
know,
maybe
proprietary,
that
they
don't
want
to
share,
and
so
that
was
something
I
had
not
thought
about
before.
But
that's
because
I've
not
worked
in
large
corporation
doing
open
source.
A
We
had
other
talks
on
some
of
the
other
aspects
of
the
project
or
the
program,
so
the
program
started
in
2005
and
it
was
very
informal
and
they've
grown
to
a
very
large
number
of
mentees
and
and
mentors
over
a
number
of
organizations
that
sponsor
different
groups,
so
we're
sponsored
by
a
ncf
and
incf
was
more
represented
there,
and
but
there
are
all
sorts
of
other
organizations
that
draw
from
that
program
from
traditional,
open
source
projects.
A
To,
I
think
wikipedia
has
some
interns.
I
think,
like
some
of
the
other
bigger
open
source
projects.
Have
some
interns
to
you
know
actual
corporations.
A
So
there's
a
very
diverse
group
of
people
now
involved
in
this,
and
so
I
I
wanted
to
share.
Also,
okay,
I
wanted
to
share
I'll
share
my
screen.
A
I
wanted
to
share
some
things
about
our.
We
have
a
onboarding
guide
here.
Actually
I
wanted
to
show
here
it
is
so
I
didn't
show
this
last
time
I
last
year
when
we
did
google
summer
code,
we
made
an
onboarding
guide
and
this
is
the
efforts
of
a
couple
of
people,
a
couple
of
former
contributors
plus
myself.
A
My
knock
deb.
I
think,
and
was
involved
with
this-
and
I
was
involved
in
this
and
trying
to
think
of
who
else
was
involved
in
this
a
couple
other
people.
So
this
is
it
kind
of
goes
through
some
of
the
basics
of
our
organization.
So
what
is
open
worm,
diva
worm
and
essential
links?
So
I've
talked
about
open
worm
before
they're
trying
to
build
the
first
digital
ice
form
in
an
open
source
fashion.
A
So
we
have
some
links
here
to
the
organization,
the
slack,
and
this
is
a
couple
of
talks.
These
are
talks
that
I
didn't
discuss
before,
but
this
kind
of
goes
through
some
of
the
open
or
materials.
A
Some
of
the
you
know
things
but
get
using
github,
although
I
don't
think
a
lot
of
people
who
I
usually,
though
we
get
applicants
who
are
different
levels
of
knowledge
of
github,
believe
it
or
not.
So
this
is
just
to
get
people
up
to
speed
on
github,
but
this
is
more
consider.
You
know
this
is
an
onboarding
guide,
more
broadly
for
people
joining
diva
worms.
So
if
they're
joining-
and
they
maybe
aren't
really
that
heavily
computational
science
oriented,
they
can
learn
about
github.
A
And
then
you
know
we
have
our
resources
for
diva
worm,
which
is
a
separate
thing
from
openworm
somewhat.
We
have
our
own
github
repository
group
website,
organization,
repository
and
so
forth,
and
then
these
are
slides.
That
kind
of
this
is
a
krishna
katyal
who
is
now
working
at
a
open
source
organization
called
or
with
an
open
source
organization
called
toki
and
they're
doing
a
lot
of
templates
for
machine
learning.
So
he
made
up
these
slides
and
they're
kind
of
fun.
They
kind
of
walk
through
open
source.
A
Why
you
should
contribute,
and
then
you
know,
giving
a
little
bit
of
information
about
our
past
achievements
and
then
getting
into?
And
then
after
that
we
have
some
general
faqs
here
and
then
some
links
to
some
of
our
model
organisms.
So
these
are
some
links
to
different
things.
If
you're
coming
into
the
organization,
you
want
to
see
what
kind
of
organisms
we
work
with
actively
so
we're
working
with
c
elegans
here,
there's
some
examples
of
c
elegans.
A
This
is
the
the
diatom
which
is
a
marine
organism.
It's
an
algae
and
they
live
in
these
colonies
and
diatoms
are
they're
actually
made
of
glass.
So
these
are
silicate
organisms
and
they
live
in
these
colonies.
Not
all
of
them
live
in
colonies,
but
the
ones
we
study
live
in
colonies,
the
ones
we've
been
publishing
on.
A
So
we
have
a
little
bit
about
diatoms
here
and
then
we
have
axolotls.
So
this
is
the
project
that
hari,
krishna
and
koran
are
working
on
and
they're
working
on,
axolotl
embryos
and
there's
a
axolotl
model
organism
community.
Just
as
there's
a
c
elegans
model
organism
community,
and
we
have
some
links
there.
Then
we
have
some
biological
data
sets.
So
there's
some
overlap
here
with
the
diva
zoo
stuff,
but
I
wanted
to
bring
this
devozu
stuff
out
and
kind
of
give
people
some
pointers
to
some
of
those
things.
A
A
Okay,
any
questions
so
far.
A
I
know
tom
hasn't
been
here
in
a
while.
Would
you
like
to
give
an
update
on
what
you've
been
up
to.
G
Oh
well,
okay,
okay,
I
just
thought
I'd
check
in
and
I
sort
of
just
finished
a
project
sort
of
got.
A
little
break
here,
worked
on
a
little
project
with
I've.
Never
seen
this
done
before.
So
I
thought
I'd.
Try
it
how
like
sequential
neural
networks,
how
well
they
perform
if
you
like,
disrupt
the
streams,
the
temporal
streams,
so
I
investigated
like
lstm
and
tcn
and,
of
course,
more
fanastic,
which
is
my
baby,
which
did
the
best
strangely
enough,
so
so
what
it?
G
What
it
means
is
like
suppose
you
have
a
task,
that's
a
stream
of
like
inputs
and
outputs
like
stimuli
and
responses,
and
so
you
learn
that
and
then
what
you're
doing
is
I
mean
this.
This
happens
to
organisms
all
the
time,
you're
going
about
your
business
and
something
happens,
and
it's
not
it's
it's.
G
You
know
you
have
a
destruction
right
and
you're,
making
a
cake-
and
you
run
out
of
you,
know
sugar
or
something
so
something
gets
injected
in
that
path
or
just
delete
it
from
the
path
and
substituted
from
the
path.
It's
a
little
piece
of
stuff.
You
know
about
it's
like
a
modular
chunk,
but
you
have
to
go
like
detour.
Do
that
and
then
come
back
to
the
original
path.
G
A
G
A
Yeah,
so
you
have
done
a
lot
of
work
with
neural
networks
and
ai.
Could
you
tell
people
a
little
bit
about
that.
G
Well,
yeah,
I
mean
you
know
mostly
you
know
doing
machine
learning
related
to
animal
behavior
like
even
as
c
elegans
can
be
modeled
as
a
kind
of
a
cellular
automaton.
G
You
know
so
I
just
I
guess
my
little
thing
is
you
know
animals
are
you
know
they
still
have
lots
of
capabilities
that
ais
don't
have
whatsoever
so
they're
very
flexible,
and
they
deal
with
changes
in
some
ways
that
are
just
amazing.
You
know
so
I
mean
I
was
just
like
reading
about.
You
know
bees
and
stuff
like
that
how
they
can
count,
and
you
know
how
they're,
how
they
remember
things,
and
they
have
this.
G
Like
group
intelligence,
that's
like
all
real
fascinating
to
me,
so
I
gotta,
I
sort
of
feel
compelled
to
try
and
get
some
of
that
stuff
and
find
out
how
some
of
that
stuff
might
be
done
in
artificial
intelligence,
and
you
know
neural
artificial
nerve.
No,
that
works.
So
you
know.
A
Yeah
yeah,
so
you
you
created
this
morphozoic
platform
and
we've
talked
about
that.
If
the
the
gsoc
people
are
looking
through
some
of
the
stuff,
I
I
sent
into
the
slack
on
the
open
worm
open
house
several
years
ago.
We
did
a
tutorial
on
on
morphozoic
and
this
is.
A
A
Okay,
yeah-
and
this
is
a
cellular
automata
that
is
these
nested
neighborhoods.
So
we
did
this
session
on
the
tutorial
and
you
know
that
software
is
still
available
for
people
if
they
want
to
use
it.
This
is
a
classic
open
source
software,
so
you
can
use
it
for
you
know.
A
If
you
have
an
idea,
I
don't
know,
maybe
it's
not
very
close
to
what
people
are
doing
in
these
projects,
but
it's
definitely
worth
you
know
revisiting,
because
I
I
we
would
been
talking
about
neural
networks
last
year
and
then
we
kind
of
dropped
it
because
we
didn't
have
well.
We
didn't
have
a
gsoc
project
this
year
on
it,
but
yeah
I
mean
we,
it's
it's
something.
It's
still
there
it's
outstanding,
so
I'm
trying
to
encourage
people.
It
never
really
got
to
it
directly,
though.
G
I
mean
you
know,
but
there's
so
many
things
happening.
You
know
so
I
mean
like
I
just
ran
across
like
within
the
last
year.
This
thing
called
the
temporal
convolutional
network.
It's
fairly
new.
It
came
out
in
2018,
so
it's
kind
of
suppose
it's
convolutional,
but
it
does
like
temporal
streams
of
you
know
things
that
are
moving
along
in
time.
So
it's
supposed
to
be
really
good.
It's
supposed
to
be
superior
in
some
some
respects
to
you
know
the
standard
like
lstms
and
then
there's
a
new
one,
which
is
a
bit.
G
I
forget
the
name
of
it,
but
anyway.
A
Yeah
we
actually,
we
just
had
a
conversation
I
think
two
weeks
ago
about
like
having
you
know,
building
networks
where
they
sort
of
repair
like
looking
at.
If
you
knock
a
part
out
and
it
repairs
itself,
you
know,
what's
the
measure
the
capacity
for
that
or
how
do
we
look
at
the
performance
of
yeah
and
they're
talking
about
neural
networks.
G
A
G
G
But
yeah
the
animals
you
know
like
I
got
involved
with
the
open
worm
project.
You
know
like
like
10
years
ago
or
something
like
that
because
just
you
know
animals
are.
I
think
it's
just
like
the
big
piece
we
gotta
try
and
figure
out
how
to
keep
working
on
that.
Get
that
into
you
know.
If
we
want
to
actually
make
ai's
real,
I
don't
you
know.
Maybe
some
people
say
we
don't
want
to
do
that,
but
you
know
like
elon
musk,
you
know
yeah,
it's
like.
Oh
no.
G
A
G
G
About
that
other
guy
I
worked
with
the
other
robot
guy
in
australia.
He's
going
great
guns.
He's
done
some
amazing
stuff.
You
know
really
yeah
sean
shane,
yeah
sean,
you
remember
him.
Oh
yeah,
yeah,
yeah,.
G
A
A
Yeah
I've
been
doing
work,
I've
been
doing
some
open
source
community
management
for
rockwire,
which
is
a
initiative
by
the
university
building
campus
apps,
and
things
like
that.
A
So
I've
been
doing
like
a
lot
of
the
stuff
that
I
I
kind
of
do
here
and
then
you
know
working
with
mobile
development
and
things
like
that.
So
it's
a
little
bit
different
than
like
open
worm,
but
it's
it's
a
interesting
thing
and
I've
been
also
been
working
with
my
own
lab
group,
where
we've
been
doing
a
lot
of
stuff
with
ai
and
and
cognition
cognitive
science.
G
A
H
G
Oh
wow,
it
looks
like
looks
like
one
of
those
fusion
things.
You
know
it's
gonna
yeah
great
great
anti-matter,
like.
H
Telescope
because
it
looks
like
a
ball,
it's
like
canadian
split,
okay,.
B
Yeah
sorry,
I
just
I've
been
trying
to
find
a
small
ball.
It
will
go
in
the
middle
of
it
and
it's
a
3d
printed.
B
G
E
E
E
E
E
H
G
E
A
Yeah,
it's
yeah.
I
don't
think
we've
gotten
anywhere
any
further
on
that.
So
that's
that's
about
it.
E
A
So
yeah,
so,
let's
see,
looks
like
taro
had
to
leave.
I'm
gonna
go
through
some
things
that
I
have
I've
been
pulling
up,
maybe
some
things
that
I've
made
we've
made
some
progress
on
so
like
I
I
think
I've
mentioned
this.
This
is
a
ways
back,
but
I
I
talked
about
the
mathematics
of
d'evil.
Oh,
I
can't
sure
there
we
go
the
mathematics
of
diva
worm,
which
was
a
sort
of
a
collection
of
diagrams.
I
was
bringing
up
and
I
don't
know
if
this
mesh
screen
is
having
trouble
sharing.
A
But
let's
see
I
can
do
this
all
right,
all
right.
There
we
go.
So
this
is
the
mathematics
of
d'evil.
I've
made
a
poster
of
it
actually,
and
this
is
kind
of
just
waiting
for
a
venue
to
to
put
the
poster
in,
but
I
can
also
put
it
up
on
on
the
web
and
we
can
have
this
as
a
standard
poster,
so
this
kind
of
goes
through
the
mathematics
of
divorm,
and
so
you
know
there
are
a
couple
of
models
that
have
been
put
out
in
papers.
A
The
developmental
function
and
spatial
complexity
and
self-organization-
these
are
some
other
things
that
are
sort
of
diagrams
kind
of
thinking
about,
like
the
basic
parts
of
this
sort
of
computational
developmental
biology
approach,
we've
kind
of
done
things
related
to
categories
and
category
theory,
which
is
diagrammed
here,
some
things
from
one
of
the
papers
on
developmental
game
theory
and
then
some
of
these
differentiation
trees
and
computational
development.
A
So
this
is
a
sort
of
a
differentiation
tree,
lineage
tree
combination
here,
so
this
is
just
to
give
people
an
idea
of
some
of
the
things
that
are
going
on.
You
know,
I
guess
I
haven't
put
the
citations
and
I'll
put
them
in
like
about
here,
but
this
is
you
know
what
I
kind
of
wanted
to
do
with
this
is
to
make
a
poster
that
kind
of
goes
over
these
different
little
methods
and
you
know,
points
people
to
papers
and
giving
them
a
visual
of
what's
going
on.
A
So
this
is
something
that,
if
you
know
we
can,
if
you're
interested
in
like
contributing
to,
we
can
talk
about
how
to
you
know,
maybe
you
know
you
know,
do
some
other
things
with
it,
but
this
is.
This
is
definitely
something
that
you
know
last
time
I
showed
it
to
the
group.
It
was
kind
of
in
these
diagrams
that
were
scattered
all
over
the
place.
A
So
if
you
look
at
these
two
figures
here
they're
these
graphs
and
these
are
actually
from
two
papers
that
we've
done
in
the
past.
Here
one
was
from
a
paper
in
2021
on
comparative
embryos,
and
this
one
was
from
like
2017
on
c
elegans,
lineage
trees
and
I'm
having
trouble
zooming
in
because
this
is
a
poster
and
it's
really
big,
but
you
can
see
these
two
figures.
A
In
any
case
I
have
in
the
meeting
here
it
is
okay,
so
this
is
kind
of
going
back
to
the
simula
or
regenerating
network
stuff.
So
in
thinking
about
this
problem
of
regenerating
networks-
and-
and
I
told
you
before
that
that
was
the
project
that
we
were
talking
about
where
or
the
idea
that
you
know
we
could
create
these
networks,
not
necessarily
neural
networks,
but
like
complex
networks
that
regenerate
their
parts.
A
If
you
take
them
off
and
regenerate
and
kind
of
measuring,
maybe
you
know
what
it
is
about:
networks
that
are
make
them
robust
or
make
them
maybe
give
them
this
property,
so
that's
kind
of
a
hard.
You
know
I'm
kind
of
approaching
this
from
some
of
the
stuff
we've
done
before
and
I'm
not
really
close
to
being,
like
you
know,
having
an
answer
for
you,
but
I
want
to
go
through
some
of
these
older
papers.
So
this
paper
was
information
isometry,
which
is
this
technique
that
we
came
up
with
in
the
group
around
2017.
A
And
then
you
can
do
things
like
flip
the
nodes
around
and
you
can
flip
the
order
and
you
can
measure
compare
different
trees
in
terms
of
their
order.
So
you
can
do
things
like
measure
different
cells,
if
they're
in
different
states
on
different
parts
of
the
tree,
if
they
get
flipped,
somehow
you
can
measure
the
distance
between
the
two
trees
and
do
other
things
now.
This
lends
itself
to
knocking
some
of
these
nodes
out
so
say
I
knocked
out
this
part
of
the
tree.
A
A
Then
you
can
actually
so
you
can
like
identify
different
defects
in
the
trees.
You
can
identify
like
things
where
trees
have
been
clipped
off
different
parts
of
the
trees,
but
then
also
you
know
if
it's
regenerating
something
it's
going
to
regenerate
according
to
some
schema,
so
you
might
start
down
here
where
it
regenerates
from
zero
zero.
One
and
there'd
be
some
additional
branching
down
here.
That
would
have
its
own
code
and
that
would
have
a
distance
from
say
like
a
regular
tree.
A
That
looks
like
this
like
a
null
tree,
and
so
you
could
actually
measure
that
regeneration
as
well.
So
there
are
ways
you
can
do
that
your
ways
you
can
encode
this,
this
removal
of
nodes
and
addition
of
nodes
and
have
a
computational
representation
of
it.
That
still
doesn't
give
you
a
mechanism,
but
one
of
the
nice
was.
E
A
E
E
Okay,
just
just
just
search
the
elegance
and
and
vogue
we
do
lva.
A
E
C
A
This
is
for
the
entire
c
elegans,
so
the
vulva
is
down
one
of
the
lineages.
This
is.
This
is
just
actually
based
on
c
elegans.
This
this
example-
and
it
just
shows
like
this
binary
division.
So
in
some
organisms
you
don't
have
this
kind
of
binary
division
where
you
have
an
identity.
There's
you
know,
proliferation
of
cells.
A
There's
you
know
they
don't
have
strict
identities.
So
this
is.
This
is
just
kind
of
like
a
c
elegans
like
tree,
but
there
are
areas
like
the
the
germline
and
the
vulva
and
things
like
that
where
the
tree
behaves
a
little
bit
differently,
but
one
of
the
cool
things
you
can
do
with
this,
then,
is
to
look
at.
A
If
you
compare
two
trees
and
you
find
the
distance
between
them,
you
can
actually
generate
random
trees
and
you
can
look
at
you
know
this
like
a
biological
tree,
so
you
can
take
a
tree
from
like
an
embryo
and
you
can
also
generate
random
trees,
which
means
you
just
generate
nodes
anywhere
and
you
compare
them
across
individuals
or
across
trees,
and
you
can
actually
get
graphs
like
this.
So
this
does
is,
it
shows
what
they
call
the
hamming
distance,
which
is
a
measure
of
the
distance
between
two
strings.
A
A
It's
very
similar
to
this,
but
the
hamming
distance
came
first
and
it's
like
they've
used
an
information
theory
for
many
years,
but
this
is
a
way
to
evaluate
binary
strings,
and
so
what
you
see
here
is
you
see
that
c
elegans
has
a
greater
hamming
distance
across
on
each
tree
level.
So
lineage
tree
level
is
how
many
branching
events
do
you
witness
from
like
a
single
cell,
so
one
a
lineage
tree
of
level.
One
has
a
certain
distance
level,
two
level
three
and
so
on,
and
so
the
deeper
the
tree
gets.
A
You
get
a
larger
hamming
distance
between
the
samples
and
you
notice,
though,
that
it
with
the
random
trees.
You
get
this
it's
much
lower
than
the
biological
case.
Why
that
is
is
well
it's
probably
because
in
the
biological
example
your
lineage
tree
is
some
structure
where
you
know
there's
there
are
different
areas
that
are
invariant
and
maybe
different
areas
that
get
turned
get
flipped.
A
So
this
is
explainable
by
just
by
virtue
of
not
having
a
fully
random
sequence,
that
certain
parts
of
the
sequence
move
around
more
than
others,
and
that
makes
sense
when
we
think
of
lineage
trees
as
being
these
things
that
sometimes
it's
important
for
things
to
be
maintained
in
them.
So
what
what
does
that
mean
about?
Like
cell
divisions,
cell
divisions,
then
you
know
were
clipping
parts
of
the
trio.
A
So
that
means
that
your
cell
divisions,
you
know,
if
you
take
out
part
of
the
tree,
you
know
you
can
represent
it
using
that
graph
and
it'll
give
you
a
representation
of.
Maybe
you
know
where
you
can
identify
these
different
places,
where
maybe
they're
more
vulnerable
to
being
cut
off,
where
they're
more
likely
to
be
regenerated,
or
something
like
that
I
mean
I
haven't,
thought
it
through
any
more
than
that.
But
that's
that's.
What
I've
been
revisiting
this
paper
and
getting
that
out
of
that
stuff?
This
is
the
other
paper.
A
This
is
the
one
that
was
published
in
biosystems
like
I
think
last
year,
and
this
one
has
a
graph
in
it
where.
A
It
basically
it's
it's
not
the
same
graph,
it's
a
similar
graph.
If
I
can
find
it
well,
I
have
it
here.
So
this
is
oh,
that's
not
it!
This
is
a
so.
This
is
a
similar
graph
where
you
have
the
number
of
divisions
versus
developmental
time.
So
this
developmental
time
in
this
graph
is
very
similar
to
the
lineage
tree
level.
A
In
this
graph,
it's
just
a
different
way
of
expressing
the
same
thing,
but
it's
over
time,
and
then
you
have
this
number
of
divisions
and
if
you
notice
this
top
blue
graph
here,
this
blue
line
is
it
looks
like
it's
just
kind
of
stepping
up.
You
know
deterministically,
and
that's
because
this
is
something
that
is
representative
of
a
lineage
tree,
that's
dividing
in
a
normal
interval,
so
it's
divided,
making
those
branches
at
a
normal
time
interval
and
it's
it's
making
a
number
of
divisions
over
time.
A
So
you
can
see
that
it's
it's
a
straight
line.
There's
nothing
unusual!
I
mean
it's
actually
unusual
in
biology
to
see
this,
but
this
is
like
a
null
tree
basically,
but
you
notice
that
if
you
use
other
distributions,
you
use
these
other
distributions
like
a
exponential
distribution
or
poisson
distribution,
and
you
can
generate
these
computationally.
These
are
biological,
but
you
get
these
different
step
lengths.
A
So
you
get
a
larger
number
of
divisions
per.
You
know,
unit
developmental
time
sometimes,
and
sometimes
you
don't
get
any
divisions
over
a
long
period
of
developmental
time.
What
will
happen
if
you
use
different
just
mathematical
distributions
is
that
you
can
stretch
out
these
steps
between
the
number
of
divisions
and
developmental
time
and
and
they're
different
ways
that
this
manifests
itself.
They
all
basically
end
up
at
a
similar
point,
but
there
are
these
different.
A
You
know
these
different
step
lengths.
So
you
know,
one
of
the
things
I'm
thinking
about
here
is
this
is
something
that
can
be
exploited
in
a
system.
That's
regenerating,
you
know.
Do
we
need
to
like
understand
some
of
these?
These
alternate
distributions
as
a
means,
for
you
know,
providing
maybe
vulnerable
a
signal
for
vulnerability
or
a
signal
for
robustness.
That
would
encourage
regeneration,
or
what
does
this
give
us
now?
This
isn't
biological,
but
what's
interesting
is
that
these
alternate
distributions
may
connect
to
some
different
types
of
biology.
A
So
if
we
build
trees
that
have
variable,
the
branching
is
variable
in
time
where
it's
different
in
different
parts
of
the
tree.
It's
variable
that
signature
then
may
tell
us
something
about
its
ability
to
regenerate,
or
we
can
run
an
algorithm
where
we
actually
take
out
parts
of
the
tree
and
regenerate
parts
of
the
tree
and
see
how
those
behave
and
then
understand
them
in
terms
of
this
sort
of
framework.
E
Question
on
that,
can
you
configure
the
bottle
where
which
is
common
in
observational
biology,
that
the
time
interval
between
divisions
gets
longer
and
longer
as
a
an
embryo
matures.
E
If
you
consider
how
I
don't
know
what
you
mean
by
developmental
time
here,
it's,
but
if
you
consider,
if
you
consider
a
normal
embryo
in
real
time,
the
time
between
divisions
gets
longer
and
longer
for
most
embryos.
A
Yeah,
it's
probably
yeah.
That's
probably
something
yeah
that
well,
you
know
yeah,
that's
something
I
take
into
account,
but
the
you
know
the
the
thing
that
I
found
interesting
about
that
was
sort
of
the
intervals
as
they're.
Not
you
know
uniform
that
they
have
this
non-uniform
property
and
that
that
might
have
some
embarrassing
role
in
like.
E
A
A
Developmental
time
is,
is
basically
real
time.
It's
just
breaking
it
up
into
parts.
So
if
you
have
like
500
minutes
of
real
time,
it's
like
so
many
units
at
developmental
time.
It's
just
another
way
of
saying
real
time,
though.
A
Well,
in
those
simulations
I
set
them
to
be
con.
Well,
I
set
the
uniform
case
to
be
constant,
but
then
the
other
simulated
cases
I
showed
that
they
were,
you
know
not
constant,
but
now
you
could
put
that
into
a
turn
that
into
a
biological
tree
and
give
it
that
property,
where
divisions
take
longer
and
longer
over
time
or
whatever
you
want
to
do.
D
G
I
really
you
know
yeah.
I
know
you've
been
working
on
this
for
quite
a
while.
I
was
just
wondering
does
when
this,
when
the
thing
you
know
subdivides
and
differentiates,
is
there
like
an
entropy
aspect
to
this,
where,
like
top
level,
divisions
tend
to
separate
into
chunks
that
are
less?
G
You
know
like
it's
like,
because
I
kind
of
bridge
things
there's
like
a
decision
tree
where
the
thing
it
reduces
the
entropy
maximally
at
the
top.
It
goes
down
like
so
like
the
sub
cells
of
the
top
level
division.
G
Do
they
give
rise
to
things
that
are
like
you
know,
the
the
complexity
of
those
sub
structures
are
maximally
plus
entropic.
You
know
that
sort
of
thing
does
it
doesn't
one
of
nature
does
that
you
know
just
eating.
A
D
A
So
they'd
be
less
subject
to
entropy
or
yeah.
Well
yeah,
if
you
take,
I
didn't
show
like
local
measurements,
so
the
hamming
distance
is
sort
of
a
measure
of
global
entropy
for
all
for
different
levels,
but
that's
across
the
entire
tree.
So
if
we
were
to
measure
some
parts
of
the
tree,
you'd
probably
get
this
sort
of
you
know
you'd
probably
get
this
sort
of
signature
of
less
entropy
over
time.
A
A
G
I
mean
we're
not
gonna
run
now
so,
okay
good
seeing
y'all.
A
Looks
like
morgan
joined
us
and
jesse
and
harry
krishna
had
to
leave.
So
thank
you
for
attending
so
morgan.
I
I
see
you
to
make
it.
A
Are
you
okay?
There
you
go,
could
you
introduce
yourself,
I
don't
think
anyone's
met
you
in
this.
Oh.
I
Yeah
hi,
I'm
morgan.
I
am
a
mirror.
I
I
I
followed
up
the
worm
base
paper
just
looking
at
the
kind
of
the
resources
and
things
that
that
are,
you
know
that
are
available,
which
are
which
are
impressive
and
and
broad
so
yeah.
It's
been
been
interesting
to
see
and
and
I've
been
trying
to
lead
a
a
brain,
organoids
cerebral
organoids
interest
group
in
the
oral
meetings
and
that's
kind
of
some
of
the
overlap
between
the
diva
worm
and-
and
you
know
things
things
I'm
interested
in.
So
it's
great
to
be.
A
Nicey
to
join
us
yeah,
we
were
just
having
a
conversation
about
some
things
in
developmental
biology
and
then
we
were
talking
about
google
summer
code
for
this
group,
which
I
know
we
have
google
summer
code
in
the
other
group.
But
in
this
group
we
have
two
projects.
We
have
the
the
graph
neural
networks
and
then
we
have
the
like
a
spherical
map
of
the
axolotl
embryo
so
that
those
are
yeah.
I
Yeah
yeah
no,
and
that's
that's
great
too
so,
like
I
said
you
know
some
of
the
overlap
here
is
that
yeah,
I
think
any
of
the
organic
work
would
would
require
a
good
sure.
I
It's
it's
my
name,
my
name
at
gmail,
but
else
I
can
send
it
to
you,
and
so
I
I
like
some
of
the
microscopy
overlap
and
you
know
really
interested
in
in
one
seeing
what
the
what's
available
here
in
the
bay
area
for
for
independent
people,
but
also
wondering
you
know
if
some
of
the
photogrammetry
work
that
we've
done
previously
might
be
useful
for
the
for
the
the
photo
stitching
that
I
think
you
were
doing.
I
You
know
yeah,
but
you
know
definitely
you
would
love
to
learn
more
about.
You
know
I.
I
know
that
chan
zuckerberg
supports
a
like
an
image
library
for
for
microscopy
yeah.
So
it's
like
something
like
n
pairs
or
something
I
I
know.
That's
actually
a
different
package.
A
I
But
it
just
it
looks
like
in
pairs,
and
you
know
so,
like
I
don't
know,
I
need
to
learn
more.
You
know
like
the
last
thing
I
I've
heard
about
is
like
image
j,
and
that
was
you
know
10
years
ago,
so
yeah
it's
been
funny
to
review
all
the
kind
of
agent-based
modeling
software,
it's
it's
like
70,
80
percent
of
it
it's
all
java
and
it's
I
swear
that's
just
because
of
the
old
swarm
packet
or
you
know,
like
probably
yeah.
There
was
this
like
java
ecosystem
anyway.
I
Yeah
really
interested
in
that
and,
of
course,
the
the
graph
neural
networks
is
it
you
know
in
in
neuroimaging
and
kind
of
connectome
analysis.
I
You
know
graph
based
analysis
is,
is
very
hot
these
days,
and
you
know
I'm
also
really
interested
in
the
overlap
between
between
areas
that
I
mean
the
kind
of
research
communities
that
that
haven't
necessarily
talked
you
know,
so
I'm
really
interested
in
the
in
our
data
without
borders,
user
days
that's
coming
up
and
really
more,
I
mean
it
would
be
interesting
to
look
at
some
of
the
file
format,
issues
and
things
like
that
from
invasive
recordings
versus
ep
recordings.
I
But
but
it's
really
looking
at
some
of
the
the
libraries
that
invasive
people
use
that
eeg
people
aren't
really
familiar
with
and
so
really
looking
forward
to
that.
A
Yeah,
so
we've
we've
actually
discussed
photo
photography
or
photogrammy
in
the
group
here,
especially
with
respect
to
this
this
project,
but
also
other
projects
and
in
our
in
our
youtube
channel.
We
have.
We
have
a
couple
of
talks
on
different
techniques
of
microscopy.
A
We
had
one
person
in
the
group
a
long
time
ago,
who's
since
deceased,
who
was
he
did
a
lot
of
stuff
with
like
he
was
an
iron
worker
and
all
these
other
things,
but
he
did
some
different
forms
of
imaging.
Like
you
know,
biological
imaging
and
things
like
that,
steve
mcgrew-
and
you
know
he
has
a
talk
in
our
one
of
our
playlists
on.
A
I
can't
remember
what
it
what
the
met
technique
was,
but
you
know
we
talked
regularly
about
these
different
imaging
techniques
and
dick,
of
course,
has
been
he's
done,
a
lot
of
work
on
this
in
his
career.
So
and
then
you
know
some
of
the
other
stuff
with
organoids.
A
I
know
that,
like
we've
talked
a
little
bit
about
organoids
in
the
group,
but
it's
been
very
you
know
it's
kind
of
like
the
problem
you
have
with
organoids
is
like
it's
very
difficult
to
get
a
sense
of
the
landscape
of
how
to
use
these
for
different
things.
So
in
this
group
we
thought
about
a
little
bit
more
broadly,
I
think
we've
discussed
like
different
organs
and
organoids
and
more
like
kind
of
how
they
integrate
with
development
than
maybe
with
like
brain
function.
But
you
know
it's
a
lot
of
really
interesting
problems.
B
I
I
Sure
that's
yeah.
I
would
love
to
see
more
about
that.
I
I
I
wanted
to
bring
out
the
the
morpheus
package
bradley.
I
don't
know
if
you
if
you've
mentioned
that
before,
but
but
you
know,
I
I
mean
in
trying
to
get
more
practical
as
well.
As
you
know,
try
and
think
of
things
that
you
know
perhaps
others
that
are
interested
to
you
know
get
involved
with
I've
been
thinking
about.
I
I
A
There
are
a
lot
of
well,
I
mean
there
are
a
lot
of
models
in
like
a
lot
of
computational
models.
Now
for
embryogenesis
I
mean
they're,
not
you
know,
they're
varying
degrees
of
usefulness
for
organoids,
but
the
idea,
you
know
that
you
can
trace
lineages
or
the
idea
that
you
can
look
at
different
sulfates.
A
You
know
and
like
organoids
aren't
the
same
as
like
biological
development,
but
you
could
get.
You
know,
take
some
of
those
principles
and
say
I'm
going
to
build
a
model
that
has
these
properties.
The
the
geometric
aspect
or
the
spherical
aspect
as
well,
is
very
important.
So
or
you
know
I
I
guess
they're,
not
really.
Some
of
them
are
spherical.
A
I
And
why
and
can
you
give
us
some
backstory
of
why
max
tegmark
is
doing
a
as
a
paper
on
3d
reconstruction
of
neurons?
I.
A
Know
well
yeah
max
tegmark
is
a
physicist
who's
done
a
lot
of
stuff
on
different
things
and
he's.
I
guess
he
put
this
paper
out.
I
put
it
in
the
slack
channel
on,
but
it's
yeah.
They.
A
I
Just
thought
yeah,
I
didn't
know
if
that's
like
an
older
research
program
that
I
just
wasn't
aware
of
or
you
know
that's
a
new
interest.
A
A
Right
right,
so
that's
something
yeah
we've!
Actually
we've
done
well,
we've
done
a
little
bit
of
stuff
with
zebrafish
data
in
the
group,
not
much
because
it's
like
I
mean
you
know
it's,
it's
a
it's
a
different
type
of
development,
different
type
of
organism,
but
it's
not.
You
know
if
you're
going
to
compare
with
other
things.
I
think
we've
also
done
some
work
with
sea
squirts
a
long
time
ago,
dick-
and
I
did
some
work
with
like
looking
at
the
data
from
c
squirts.
A
So
some
of
the
papers-
I
was
just
showing
you,
you
know
they
are
sort
of
like
have
this
comparative
aspect
where
you
have
two
different
species,
they're
a
very
different
development,
but
you
can
still
gain
insight
from
that
because
they
have,
you
know
they
have
processes
of
cell
division.
They
have
processes
of
cell
differentiation.
A
Those
are
things
you
can
compare
and
say
you
know
there
are
these
things
that
they
have
in
common
these
things
they
don't
have
in
common.
So
it's
it's
a
you
know
it's
nice
to
have
that
comparative
approach.
So
it's
very
much
very
useful
and
I
think
you
could
do
the
same
with
organoids.
E
A
I
A
A
Yeah,
bye,
okay,
well,
have
a
good
week.
Everyone
all
right,
bye,
hello.
Now
I
want
to
go
over
a
paper
on
body
plan
identity.
This
is
something
that
just
came
out
in
the
literature
and
hopefully
it's
informative
to
some
of
our
students
and
it'll,
maybe
foster
some
conversation
amongst
our
other
collaborators.
A
So
this
is
a
vapor
body
plan
identity,
mechanistic
model-
this
is
was
published
in
evolutionary
biology.
I
believe
this
year
and
it
was
march
21st
and
it's
by
james
defresco
and
gunter
wagner.
Winter
wagner
has
done
a
lot
of
stuff
with
theoretical
work
on
robustness
and
evolvability.
So
this
is
an
interesting
perspective
on
body
plan,
identity
and
development.
A
So,
let's
zoom
in
on
the
abstract
here
and
let's
go
through
this
paper.
So
the
abstract
reads:
a
body
plan
is
a
stable
configuration
of
characters
for
a
major
taxonomic
group
such
as
chordates
or
arthropods.
So
a
body
plan
is,
of
course,
the
sort
of
architecture
of
an
organism
and
in
evolution
we
see
a
number
of
body.
Plans
emerge
quite
early
actually
in
the
cambrian
period
and
those
body
plans
have
remained
consistent
throughout
evolution.
A
So
you
get
a
number
of
body
plans
that
emerge
in
that
several
million
years
ago,
100
million
years
ago,
and
continue
on
through
today
with
some
modifications,
so,
for
example,
arthropods,
which
are
these
segmented
invertebrate
organisms
like
different
insects,
different
crustaceans
organisms
like
that
they
have
this
segmented
body
plan.
You
see
this
again
and
again
across
different
species
of
that
taxonomic
group,
chordates,
being
organisms
with
a
spinal
cord
or
with
a
notochord
or
with
other
types
of
central
cords,
nervous
nerve
cords.
A
They
all
have
the
same
body
plan
where
they
have
bilateral
symmetry
in
this
nerve
cord,
and
so
what
they're
saying
is
that
this
is
a
stable
configuration
of
characters,
meaning
that
it's
a
stable
sort
of
configuration
of
traits
and
how
they're
arranged
in
space.
So
this
is
the
approach
they're
taking
to
development.
It's
very
evo
devo,
despite
widespread
causal
reliance
on
the
concept
for
guiding
comparisons
within
and
between
groups.
The
nature
of
body
plans,
as
well
as
the
biological
causes
underlying
the
revolution,
have
remained
elusive.
A
So
we
don't
necessarily
know
why
they've
evolved
in
the
way
that
they
did
and
we
don't
know
why
they
remain
stable.
We
just
know
that
they
do
so
it's
it's
not
really
easy
to
figure
that
out.
We
just
kind
of
make
this
assumption
that
I
mean
we
can
see
in
the
data
that
it's
remain
stable.
We
don't
exactly
know
why,
other
than
to
say
it's
been
conserved,
a
lot
of
people
say
it's
conserved
in
the
sense
that
it
doesn't
change,
but
why
doesn't
it
change?
A
This
paper
proposes
an
abstract
mechanism,
mechanistic
model
of
body
plan
identity,
so
they
go
into
this
idea
of
positional
identity
or
a
certain
type
of
developmental
identity,
which
requires
the
cells
to
have
an
identity
that
is
comparable
between
cells
and
that
puts
them
in
a
certain
order.
So
we've
talked
about
in
different
group
meetings.
We've
talked
about
positional
identity
and
how
cells
have
an
idea
of
kind
of
where
they
belong
in
the
embryo,
where
they
should
migrate
to
and
what
they
should
do.
A
We
hypothesize
that
body
plans
are
an
evolutionary
phenomena
that
only
applies
to
a
relatively
small
subset
of
major
clades
and
clades
and
revolutionary
groups
in
a
tree
rather
than
being
associated
with
each
and
every
so-called
phylum.
And
so
this
is
you
know
where
you
have
different
body
plans
and
different
parts
of
the
tree.
They
don't
necessarily
map
to
our
our
nomenclature
that
we
use
in
science.
A
They
map
to
organisms
that
descend
from
a
common
ancestor
and
whether
those
be
in
any
defined
phylum
or
not,
which
is
just
the
name
that
we
give
a
group,
taxonomic
group,
isn't
really
the
point
body
plans
arise
in
evolution
by
stepwise
accretion
and
require
a
level
of
developmental
complexity.
That
is
only
found
in
some
animal
glades.
A
Further.
We
suggest
that,
parallel
to
the
developmental
mechanisms
controlling
character,
identity,
there
are
body
plan,
identity
mechanisms
or
bpims
that
maintain
entire
configurations
of
characters
while
possessing
a
mechanistic
architecture.
That
itself
is
itself
stable
and
traceable
through
evolutionary
change.
So
this
is
again
this
idea
of
these
mechanisms
that
enforce
body
plan
identity.
A
They
maintain
these
different
configurations,
so
they
maintain
different
segments
of
the
organism.
They
maintain
this
bilateral
symmetry
and
they
do
this
in
a
way,
that's
consistent
and
that
you
can
trace
the
revolution
and
that's
stable
through
evolution.
A
A
So
again,
this
is
something
this
is
they
hypothesize
that
they're
part
of
these
intracellular
signaling
networks?
So
a
signaling
network
is
something
that
allows
cells
to
intracellular
networks,
allow
cells
to
communicate
with
one
another.
They
secrete
different
signaling
hormones,
they
secrete
different
types
of
signals.
Chemical
signals,
electrical
signals,
other
things
that
allow
them
to
communicate
with
their
neighboring
cells
and
when
cells
exist
in
a
group
of
cells
together
in
a
certain
location,
they
can
take
on
a
certain
fate.
A
So
if
you
take
a
stem
cell,
when
you
put
it
in
the
middle
of
a
bunch
of
liver
cells,
that
cell
is
going
to
become
a
liver
cell
where
it'll
die,
and
so
that
enforcing
differentiation
mechanism
is
actually
quite
important
to
maintaining
the
integrity
of
an
organ.
If
a
stem
cell
is
dropped
in
the
middle
of
a
bunch
of
liver
cells
and
doesn't
become
a
liver
cell
or
doesn't.
C
A
A
The
activity
of
a
bpin
results
in
a
transient
long-range
integration
of
the
embryo
that
is
highly
sensitive
to
genetic
and
environmental
perturbations,
and
that
can
be
detected
morphologically
as
a
conserved
phylotypic
stage.
So
the
phylotypic
stage
is
a
part
of
development
where
everything
is
really
conserved
and
that
you
get
the
phylotypic
meaning
it's
that
phylotype.
A
It's
this
specific
type,
that
the
embryo
conforms
to
it's
a
period
of
development
where
there's
a
lot
of
general
conformity
and
then
later
on,
you
get
this.
It's
that
people
have
talked
about
the
hourglass
of
the
phylotypic
stage,
the
phylotypic
stage
being
at
the
center
of
the
hourglass,
the
thin
part
of
the
hourglass,
the
early
part
of
development
being
beneath
that
in
the
later
part
of
development.
A
Being
above
that,
and
so
this
phylotypic
stage
can
sort
of
enforces
a
lot
of
these
mechanisms,
and
it
also
allows
the
embryo
to
be
sort
of
robust
to
some
of
these
environmental
and
genetic
perturbations.
A
So
these
bpims
are,
you
can
see
these
invertebrates
and
arthropods
respectively.
You
see
notochord
signaling,
invertebrates
segment,
polarity
and
arthropods.
A
We
conclude
by
contrasting
the
proposed
developmental
mechanistic
conception
of
body
plans
with
the
phylogenetic
notion
of
ground
plans,
so
they're
introducing
a
new
piece
of
jargon
here.
Instead
of
body
plans,
you're
using
the
term
ground
plans,
and
so
this
just
kind
of
brings
it
more
towards
a
phylogenetic
interpretation
rather
than
a
developmental
mechanistic
interpretation
and
sketch
the
general
outlines
of
an
empirical
research
program
body
plan
evolution.
A
So
this
is
so
they
kind
of
re
review
the
problem
here
in
the
introduction
they
talk
about
some
of
these,
so
the
origins
of
of
the
body
plan
as
a
theoretical
construct
and
as
an
empirical
construct.
A
So
a
curious,
evolutionary
feature
of
body
plans
is
their
entrenchment
once
they
involve
further
evolutionary
change
seems
to
be
largely
restricted
to
modifications
of
that
basic
design
rather
than
transformation
in
a
novel
or
existing
body
plan,
except
in
cases
a
major
reduction
in
body
complexity
such
as
in
parasitism
or
some
other.
A
Situation
like
that,
so
this
striking
phenomenon
is
classically
framed
as
a
problem
of
understanding
why
the
major
major
animal
body
plans
have
remained
largely
static
since
the
cambrian.
So
this
is
again
hundreds
of
millions
of
years
ago.
We
get
these
body
plans
and
then
we
stop
evolving
body
plans.
So
why
don't?
We
just
evolve
indefinite
body
plans,
and
the
answer,
of
course,
is
not
necessarily
well
known.
You
know
we're
trying
to
figure
out
exactly
why
that
is.
A
Some
people
think
that
there
have
been
novel
body
plans
that
have
appeared
since
the
cambrian
homo
metabolism
larvae
being
an
example
of
that.
So
this
is
in
insects,
there's
a
maybe
an
example
of
this.
In
any
case,
there
is
much
more
to
the
concept
of
a
than
a
puzzling
evolutionary
pattern
to
be
explained,
so
this.
A
They
talk
about
sort
of
the
details
of
the
body.
The
concept
of
body
plans
some
of
this
mechanism,
so
with
this
end,
all
right,
so
they
talk
about
the
positive
role,
for
the
concept
is
significantly
enhanced
by
mechanistic
models
of
the
interdependencies
and
development
that
give
rise
to
evolutionary
conservation
of
body
plans.
A
With
this
and
in
view,
we
propose
a
conceptual
model
of
body
plan,
identity
mechanisms
which
are
those
bpims
building
on
the
framework
of
character,
identity
mechanisms.
So
this
is
something
that
they've
worked
on
previously
in
more
of
a
phylogenetic
context,
using
characters
or
characters
being
traits
that
evolve.
You
can
identify
a
basal
version
of
it
in
a
metaphor
or
a
derived
version
of
it.
The
derived
version
is
usually
what
it's
evolving
towards
in
different
lineages.
A
So
you
have.
You
have
many
derived
forms
and
a
single
basal
form
usually
so
this
is
the
kind
of
the
way
they're
kind
of
approaching
this,
together
with
work
on
the
the
developmental
hourglass
model
from
rudy
raff,
who
is
a
famous
developmental
biologist,
and
this
hourglass
model
is
the
one
I
I
mentioned
where
you
have
this
phylotypic
stage
in
the
middle,
you
have.
A
The
phylotypic
such
as
part
of
the
hourglass,
is
where
things
are
largely
conserved,
and
this
is
somehow
related
to
this
generation
of
body
plans
and
development,
so
they
actually
propose
two
models:
they
have
a
developmental
mechanistic
body
plan
concept
and
the
phylogenetic
concept,
and
so,
if
you
take
that
extra
phylogenetic
view,
you
see
that
you
can
actually
get
a
little
bit
more
out
of
this
concept
than
you,
otherwise
would
so
again
body
point
identity,
identity
versus
state.
They
talk
about
this.
A
They
talk
about
the
concept
of
homology,
which
are
shared
derived
characteristics
or
characters
and
following
the
parallel
with
homology,
we
propose
that
there
is
a
biologically
meaningful
distinction
between
body
plan,
identity
and
body
plan
state.
So
these
are
two
things:
identity
and
state,
which
means
that
you're
simply
there's
an
identity
which
is
where
you
know,
there's
like
a
it's
like.
You
belong
to
some
group
versus
a
state.
Are
you
in
that
group
currently?
A
And
so
there's
this
and
there's
this
discussion
by
gunter,
wagner
on
character,
identity
and
character,
state
in
evolution.
So
there
are
these
phenotypic
characters
that
they're
they're
trying
to
sort
of
understand
the
state
versus
the
identity.
So
you
know
you're
identifying
a
character
versus
having
a
state
of
a
character,
so
we
can
identify
something
as
having
a
body
plan
right,
whether
it
has
a
certain
body
plan
or
not,
and
then
whether
that
organism
actually
is
in
the
state
of
that
body
plan.
A
So
you
know
there's
a
lot
of
variation
in
body
plans,
although
you
can
identify
them
by
looking
at
different
organisms
that
have
them
but
they're,
not
always
in
the
same
state.
And
so
it's
there's.
This
there's,
you
have
to
distinguish
between
both
of
them,
so
the
character
identity
of
the
tetrapod
forelimb,
for
example,
refers
to
its
identity
as
a
body
part
which
can
exist
in
many
different
character,
states
in
particular
tetrapod
lineages.
A
So
you
can
have
wings
legs,
flippers,
they're,
all
basically,
four
limbs
and
those
are
states
of
four
limbs,
but
the
actual
identity
of
a
four
limb
persists
across
all
of
them.
So
you
can
have
different
forms
emerge
and
you
can
still
identify
identify
it
character.
Identity
is
based
on
the
relative
position
of
the
character
within
the
body
plan,
as
well
as
the
other
operational
criteria
traditionally
used
to
establish
homology.
A
So
they
get
into
this
long
discussion
of
different.
You
know
mapping
this
sort
of
a
phylogenetic
approach.
You
know
you're
talking
about
positional
identity
in
some
cases,
but
you're
also
talking
about
putting
positional
identity
in
the
context
of
these
different
different
parts
of
the
body
plan.
So,
for
example,
you
know
a
trait
or
a
cell
must
be
in
a
certain
place,
so
the
four
alone
must
be
towards
the
front
or
the
anterior
end
of
the
organism.
A
So
the
cells
have
to
have
positional
identity,
to
know
that
they're
going
towards
they're
going
to
be
part
of
the
four
limb,
but
then
that
foreign
itself
can
take
on
different
forms.
It
can
be
a
wing,
it
can
be
a
flipper
and
that's
part
of
the
sort
of
specialization
locally
of
the
cells.
You
know
what
do
they
become?
What
is
their
intercellular
signaling
tell
them
to
become
so
they
kind
of
go
through
some
of
the
history
of
this.
A
The
body
plan
is
not
the
same
as
an
idealistic
archetype
concept,
which
is
something
that's
been
around
for
since
pre-darwinian
times.
This
is
but
the
body
plan
is
a
much
more
refined
concept,
but
it's
still
not
you
know,
even
though
it's
post-darwinian
and
it's
not
really
been
incorporated
with
phytogenetics,
so
they're
kind
of
done
here,
they're
identifying
mechanisms
they
identify
these
character,
identity
networks,
which
is
something
that
gunter
wagner
came
up
with
in
the
past.
A
This
involves
gene
regulatory
control
of
character,
identity.
So
now
we
get
into
gene
regulatory
networks
and
of
course
this
you
know
this
links
to
the
hox
gene,
the
collinearity
of
hox
genes
and
other
genes
like
ebx,
which
controls
high
wing,
character,
identity
and
insects.
A
Those
genes
then
exist
in
networks,
and
you
can
knock
down
certain
genes
and
show
that
they
have
developmental
defects,
but
you
can
also
show
that
they
modify
that
body
plan
a
little
bit.
Actually
they
don't
necessarily
modify
the
body
plan.
They
change
some
of
the
traits
in
that
body
plan,
so
they
they
do.
There
is
this
aspect
of
sort
of
the
modifying
the
character
state,
but
you
do
have
this.
You
still
have
the
basic
body
plan,
even
if
you
do
something
like
that.
A
So
it's
a
very
interesting
relationship,
the
these,
these
sorts
of
studies
with
different
genes
knocking
out
genes
and
showing
that
they're,
you
know
additions
or
deletions
in
the
phenotype.
Those
are
classic
evo,
devo
experiments
and
linking
those
to
this
sort
of
phylogenetic
approach,
I
think,
is
quite
valuable.
A
And
they
kind
of
go
through
these
character,
identity
mechanisms
for
anatomical
units
at
different
levels
of
organization.
So
for
cell
types
you
have
gene
networks,
transcription
factor,
complexes,
non-coding
rna
for
tissue
types.
You
have
cell
types
and
extracellular
matrix
and
for
organs.
You
have
signaling
molecules
and
signaling
centers,
so
these
are
different
ways
that
you
can
identify
these
character,
identity
mechanisms
at
different
levels
of
organization-
and
you
know
you
can
make
that
connection
between
that
and
the
phenotype
in
the
body
plan.
A
So
I'm
looking
to
see
if
there
are
any
nice
figures
in
here
and
I
was
kind
of
hoping
for
a
diagram
or
two
here's-
a
diagram
of
key
components
of
the
notochord
signaling
system.
So
this
shows
the
notochord
as
it's
forming
and
the
different
tissues
and
how
they're
all
sort
of
in
this
in
this
in
this
sort
of
arrangement-
and
this
doesn't
vary
too
much
across
species,
so
things
have
to
be
in
a
certain
anatomical
location,
and
then
you
have
aside
from
this
anatomical
organization-
that's
conserved!
A
You
have
these
different
signal
ingredients,
so
you
can
see
that
there's
signal
ingredients
that
set
up
these
different
parts
of
the
body,
so
the
bmp
gradient
and
the
wnt
gradients
start
at
the
anterior
end
and
fade
towards
the
posterior
and
shhh
gradient
is
both
sort
of
anterior
posterior
and
towards
the
middle.
You
get
a
higher
concentration.
A
This
body
plan
available
to
it
so
another,
let's
see
if
there
are
any
other
figures
here,
there's
another
figure
where
we
show
the
segmentation
body
plan.
So
this
is
a
different.
This
is
in
arthropods.
This
is
where
you
have
this
so
simplified,
simplified
illustration
of
the
spn
showing
the
basic
position,
positive
feedback
circuit.
A
So
this
is
spn,
I'm
not
sure.
I
can't.
I
don't
think
we
encountered
that
that
abbreviation
before,
but
that's
this
shows
basically
the
formation
of
these
parasegmental
boundaries.
So
you
see
in
in
the
arthropody
of
these
different
segments
and
in
between
the
segments
you
have
these
parasites
or
these
parasegments
that
show
up
in
the
body
plan,
and
these
are
repeated
across
the
body
plan
anterior
to
posteriorly,
and
these
are
small
regulatory
networks
that
have
this
motif,
so
they
always
have
the
same
sort
of
motif.
A
They
regulate
each
other
in
a
certain
way
and
they
regulate
themselves
across
the
segments.
So
each
segment
can
has
has
the
same
effect.
They
have
an
internal
regulatory
network
and
they
have
this
intracellular
regulatory
effect,
so
you
have
hh
and
wg
and
they're
regulating
their
neighboring
segments
and
they're.
Basically,
setting
up
this
gradient
across
the
segments
and
communicating
with
each
other
to
make
sure
they're
in
the
right
position,
so
there's
a
lot
of
yeah,
so
they
go
through
this
model
as
well.
A
A
Their
central
hypothesis
is
that
these
bpims
have
the
same
basic
architecture
and
causal
profile
as
character,
identity
mechanisms
for
organs.
So
these
are
cell
cell
signaling
networks,
characterized
by
active
modularity,
which
means
components
that
are
distinct,
complex
organization,
collective
necessity
and
non-redundancy
in
the
bpims.
However,
the
tissue
domains
covered
by
these
signaling
mechanisms
are
weakly
individualized
transient
embryonic
structures,
often
distinct
from
germ
layers
that
are
destined
to
differentiate
into
distinct,
individualized
characters
and
organs
under
control
of
different
ch
ioms.
A
So
this
is
so
that's
all
I
wanted
to
talk
about
with
this
paper.
Hopefully
this
was
educational.
If
you
have
any
other
questions
on
that,
please
let
me
know
thanks.