►
From YouTube: DevoWorm (2023, Meeting #16): SAM + DevoLearn, Miller-Urey + RNA World, Mammalian Cells, C. elegans
Description
Segment Anything Model (SAM) and integrating multiple forms of cell segmentation into DevoLearn. Early Life, RNA World, and the Miller-Urey experiment. Spontaneous, ubiquitous Biochemistry. Design principles for Synthetic Mammalian Cells. New C. elegans papers on larval connectomics and low-dimensional developmental manifolds. New Attendees: Susan Crawford-Young, Jesse Parent, Richard Gordon, Sushmanth Reddy, Jyothi Swaroop, and Bradly Alicea
B
A
A
B
A
B
B
D
B
D
B
There's
been
a
lot
of
work
done
here,
so
this
is
Diva
github.io.
This
is
the
Lesser
known
website,
it's
kind
of
like
the
main
website
at
weebly.com,
and
then
this
website,
which
has
a
lot
of
specialized
things
on
it.
So
there's
Devo,
Zoo,
there's
Devo
notes,
Devo
methods,
Divo
U,
Vivo
or
ml
B,
warm
AI,
open,
Devo
cell
Diva
worm
blogs
divorce.
So
all
these
are
like
these
kind
of
categories
or
Links
at
the
top.
B
Some
of
I
mean
I
think
this
probably
needs
to
be
rethought
anyways.
But
we
did
this
Diva
worm
ml
course
several
years
ago,
and
we
have
the
materials
here:
evil
or
mu.
I,
don't
know
if
that's
gonna
continue
on
Devo
methods
is
like
a
catalog
of
methods
that
we
have
Devo
notes
is
kind
of
like
methods,
notes
and
then
Devo
zoo
is
this.
So
this
is
Devo
Zoo.
This
is
what
the
idea
here
behind
this
is.
We
have
a
bunch
of
model
organisms
that
we
talk
about.
B
B
The
thing
I
wanted
to
do
with
this
is
maybe
provide
a
little
bit
more
first
of
all,
maybe
more
model
organisms,
but
also
more
just
kind
of
more
description
of
the
model
organism
and
some
of
the
things
that
are
make
them
distinct,
because
I've
seen
a
lot
of
people.
You
know
have
asked
me
like
questions
about.
B
You
know,
you
know
General
references
in
here
and
then
the
the
data
sets,
and
then
you
know
it
just
gives
people
an
opportunity
to
sort
of
like
a
Clearinghouse
for
different
model
organisms,
working
with
different
model
organisms
and
so
forth.
The
reason
I
wanted
to
redesign
it
is
because,
like
I
said,
I
wanted
to
update
it.
I
wanted
to
provide
new
model
organisms
if
we.
B
B
B
B
A
D
B
B
C
A
C
B
So
this
is
the
Devo
zoo
that
was
the
Devo
zoo
and
actually
geotheon
search
month
have
been
doing
some
good
work,
putting
like
devorn
blogs
up,
for
example,
we
have
the
blog
posts,
so
we
have
a
lot
of
blog
posts
that
we've
that
I've
done
on
my
blog
and
some
other
things
that
are
more
General
updates
and
they've.
Oh.
D
B
Yeah,
so
this
is
23
21
20.,
and
this
is
the
the
gsoc
work
from
my
knock,
my
yoke.
So
back
then
they
had
to
do
these
weekly
blog
posts
for
gsoc.
So
these
are
the
weekly
blog
posts
and
they're
kind
of
things
that
can
point
people
to
to
sort
of
explain
different
aspects
of
Divo
learn
yeah.
So
this
is
just
kind
of
going
through
this
and
then
this
is.
These
are
deworm
blog
posts.
Actually,
this
is
from
20
well
from
all
years,
so
this
goes
back
to
like
2014.
B
A
D
D
D
D
Some
of
some
of
the
data
is
missing,
I
think,
but
what
it
is
working
fine
here
collapse
itself
also
is
segmenting
properly.
I
can
be
I
think
we
can
use
it,
for
we
have
three
segmentation.
Modems
one
was
myops,
one
was
metas
Facebook
and
one
was
my
my
idea,
which
was
the
one
which
I
was
planning
to
make.
We
can
see
which
one
works
fine
and
we
can
integrate
with
demograph
with
them.
That's
what
I
was
thinking
almost
this
one
for
meta.
D
I
just
need
to
create
a
point,
develop
and
send
my
complete
code
in
it
and
that's
what
it
and
the
segmentation
model
is
perfectly
fine.
We
can
easily
integrate
it
with
their
own
and
we
can
work
with
video
data
also
and-
and
it's
not
only
working
on
sale
again,
so
it
is
working
on
different
cells.
Actually,
there
were
one
works
on
different
sense:
I
think
I
tested
with
the
emailing
star
and
more
cells
also
is
segmenting
properly,
so
we
can
create
a
sun
segmentation
model
integrated
with
metas
Facebook
and
we
can
host
it.
D
D
D
D
D
B
Yeah
I
was
really
interested
in
that
discussion
in
the
slack
channel
we
had.
Of
course
we
had
a
Manchu
put
in
the
toolkit
and
I
didn't
really
get
a
good
sense
of
I
mean
you
know
it's
what
it
can
be
used
for.
I
mean
I,
know
it's
useful,
but
I,
don't
know
how
it
would
fit
into
the
overall
scheme
of
things.
B
But
then
you
know
we
have
this
conversation
about
the
segment
anything
model
and
I
think
it
was
jiahangla
mentioned
that
you
know
be
kind
of
an
interesting
way,
I
think
to
get
into
some
of
the
graph
representations
as
well
so
yeah
I
mean
you
know.
It's
it'll
be
interesting
to
see
how
that
fits
together,
but.
B
This,
oh,
this
was
the
where,
from
where,
like
the
jiahang
was
talking
about
how
you
know
it
could
be
used
in
in
terms
of
the
graph
representations
and
then
camancho
I
think
provided
the
link
to
the
ttk
toolkit,
which
I
hadn't
really
considered
how
it
could
fit
into
the
overall
scheme
here.
But
I
think
it'll
I
think
it'll
be
an
interesting
Edition.
D
Github
page
I
think
so
it
is
nothing
is
that
just
some
redmi
files
are
there,
but
there
is
no
code
in
it
and
then
it
is
a
it's
something
like
some
startup
work
to
apply
to
control
it.
He
was
working
with
some
new,
some
Professor
I
think
so
it's
not
a
toolkit.
Actually,
the
toolkit
which
you
have
shared
was
the
perfect
one
to
get
topological
data
analysis.
Okay,
yeah
I
will
provide
the
links
we
can
use
it
perfectly.
Actually,
I
have
run
through
the
tests.
B
B
There's
they're
tectonic
puzzles
time
tracking,
uncertain
starting
vortex,
whereas
persistence
more
smelly,
quadrangulation,
so
yeah
I
mean
these
are
just
like
examples.
I
guess
where
they.
This
is
the
tectonic
puzzle
where
it's
you
know
it's
like
the
surface
of
a
planet
when
you're
trying
to
find
the
fault
lines.
Actually
that
would
be
a
good
thing
to
look
at
for
certain
aspects
of
development.
B
B
B
I
could
say
a
python
code
here
for
some
of
them,
I,
don't
know
you
know
it's
always
a
challenge
to
like
fit
the
python
code
into
what
you're
trying
to
do
so,
but
yeah.
They
have
different
examples
and
a
very
generic
but
I
think
we
can
use
a
lot
of
this,
or
at
least
some
of
it
I
mean
you
know
we
have
to
be
kind
of
targeted,
but
yeah
I
think
that's
a
really
good
set
of.
D
But
we
can
extract
the
topological
data
analysis
and
cross
verify.
Actually,
this
is
just
not
giving
the
right
that
time
or
all
these
things,
but
it
is
also
feeding
the
instant
segmentation
of
the
cells,
so
we
can
cross
verify
our
development
model
and
this
one
to
extract
you
know
cross
verify
it
I
mean
this
toolkit
expressionally
this
one
examples.
It
is
mainly
meant
for
microscopic
image,
and
this
is
the
persistent
diagram
clustering.
This
is
wrong.
D
This
is
meant
for
some,
but
this
one,
which
is
there
it
is
meant
for
microscope
image,
is
to
extract
those
kind
of
data.
Actually,
everything
is
different.
It
is
not
related
to
each
other
like
a
tectonic
plate
screen
is
like
it
will
give
you
something:
data
related
to
X
scanning
images.
It
will
generate
the
topological
data
analysis
of
that
yeah.
If
we
give
here
microscope
EMAs
here,
it
will
generate
the
proper
TDA
data
for
display.
So
that's
it.
That
is.
C
C
A
A
A
A
B
So
I
think
the
other
thing
that
I
wanted
to
bring
up
about
the
slack
channel
was
that
I
think
geohan
also
mentioned
that
there
was
that
last
year
and
when
they
were
doing
the
graph
neural
network
stuff,
they
were
trying
to
figure
out
how
to
extract
video
from
or
from
the
movies.
So
you
know,
each
frame
is
a
certain
amount
of
time,
but
of
course
there's
the
the.
B
When
you
get
the
raw
data,
it's
not
full
perfectly
aligned
with
any
one
part
of
development,
so
you
have
to
kind
of
figure
out
what
the
time
flow
is.
I
mean
it's
evenly
sampled,
but
you
have
to
figure
out
where
it
fits
in,
but
he
was
saying
that,
like
some
of
the
methods
don't
take
each
image,
they
don't
take
each
frame
independently
and
so
I
I,
don't
I,
don't
know
I,
don't
have
the
comment
in
front
of
me,
but
it
was
an
interesting
thing
probably
to
follow
up
on
in
terms
of
analyzing
movies.
B
So
you
know
you
have
I,
don't
know
what,
and
sometimes
you
don't
have
to
take-
take
it
frame
by
frame
for
a
movie
if
you,
if
you
want
to
take
a
different
sampling
strategy
for
something
like
that,
it's
perfectly
okay,
it
just
you
know,
you're
gonna
have
to
think
about
sort
of
what
the
what
the
data
look
like.
You
know,
if
there's
a
lot
of
change,
if
there's
very
little
change
going
on
for
a
number
of
frames.
B
So
that's
something
you
think
about
too
with
respect
to
analyzing
like
movie
data,
so
you
have
multiple
frames
and
having
a
Time
series
of
development.
So
I
guess
the
you
know.
The
one
reason
you
might
want
to
do
that
instead
of
taking
every
frame
is
because
you
want
to
minimize
the
amount
of
computational
work
done.
You
know
and
kind
of
maximize
the
information
you're
extracting.
B
So
I
mean
you
know,
that's
something
else.
You
might
talk
about
too
I,
it's
probably
a
topic,
maybe
for
later
in
the
in
the
summer,
as
we
work
on
this
and
a
word
about
that,
so
gsoc
results
will
come
out
tomorrow,
so
I
think
tomorrow
we
we
I,
don't
know
how
many
slots
we'll
get
that's
determined
by
Google
and
with
incf,
so
they
have
to
have
slots
available
and
then
incf
you
know,
has
a
number
of
projects
other
than
ours
that
there'll
be
people
will
be
competing
for
slots.
So.
C
B
Don't
know
how
many
we'll
get
maybe
we'll
get
all
of
them.
Probably
not
I
mean
you
know.
Last
year
we
got
a
slot
for
everyone
who
applied,
but
that's
not
probably
not
going
to
happen
this
year.
We'll
get
you
know.
Hopefully
a
couple
of
slots
and
we'll
see,
and
so
I
think
that
you'll
just
be
getting
an
email
from
I.
Think
Google
and
they'll
tell
you
whether
you've
been
accepted
or
not.
B
So
that's,
that's
all
there
I
just
want
to
make
sure
people
were
aware
of
that,
and
so
then,
if
once
you
get
your
acceptance
and
even
if
you
don't
get
your
acceptance,
we'll
have
we'll
start
our
summer
session,
so
the
community
period
starts
I,
guess
as
soon
as
you
get
accepted.
So
you
know
the
community
period
is
where
we're
going
to
introduce
you
to
the
larger
community.
B
We're
going
to
have,
you
know,
give
you
opportunities
to
learn
about
open
source,
and
then
you
know
you
can
get
started
on
sort
of
laying
the
groundwork
for
your
project.
Your
coding
period,
work
and
so
I.
Think
sushma,
for
example,
has
been
doing
that
very
well,
and
so
this
is
all
yeah.
Well,
geograph
in
for
this
here,
do
you
suck
or
are
there
more
phases
to
it?
Well,
we'll
be
continuing
with
so
devograph
and
Devo
learner
part
of
the
same
project
and
we'll
be
continuing
with
that.
B
B
We
have
you
know,
opportunities.
If
you're
not
accepted
into
gsoc
formally,
we
would
have
opportunities
to
join
in
and
work
on.
You
know
if
you
had
written
a
proposal,
you
can
work
on
that.
If
not,
you
know,
you're
welcome
to
work
on
things
and
we'll
continue
to
work
over
the
summer.
So
you
know
the
reason
I
invite
people
who
haven't
been
accepted-
or
you
know
maybe
haven't
done
a
application-
is
because
we'll
have
to
be
very
active
during
that
period,
so
you'll
be
able
to
interact
with
people
working
on
different
issues
and
different.
B
You
know
things
that
they're
they're
actually
actively
working
on
and
committing,
and
so
that's
that's.
It
I
want
to
have
like
a
working
group,
basically
around
that
over
the
summer,
so
we'll
be
meeting
at
this
time
every
week,
and
you
know
this.
This
will
allow
us
to
have
reports
on
progress
and
you
know,
go
over
topics
and
things
like
that,
so
we're
getting
ready
for
gsoc
and
Jesse
is
helping
me
with
some
of
the
logistics
of
that
the
commenter
I'll
be
co-mentoring.
B
This
project
with
Gia
Hong
and
mayuk
they've
they'll
be
I.
Guess
you
know,
advising
on
things,
they'll
be
helping
people
in
the
slack
or
in
the
GitHub
repositories,
so
such
month
of
course,
is
interacted
with
both
of
them.
So
you
know
that
that's
good,
but
those
will
be
our
our
other
co-mentors
and
then
Jesse
is
co-mentoring
on
another
project
that
we
are
proposed.
B
B
Okay,
yeah,
you
can
just
put
something
in
the
chat.
Anyways
I
have
a
couple
of
things
to
talk
about.
Today
we
have
I
wanted
to.
Actually
I
was
going
over
the
paper
that
dick
and
I
are
working
on
with
another
co-author,
and
it's
really
oh,
will
we
at
least
get
a
single
slot
this
year,
I
think
yeah.
We
should
get
a
single
slot,
I
hope
so,
but
yeah
we'll
find
out
tomorrow,
I
guess
we
usually
get
one
slot
at
least
so.
B
B
So
this
article
is
from
Scientific
American
Roger,
Shapiro
I
think
published
a
lot
of
stuff
on
this
earlier
or
aside
from
this
in
in
scientific
journals,
but
this
is
from
2007
and
this
article
is
a
simpler
origin
for
life.
So
this
the
subtitle
here
is
the
sudden
appearance
of
a
large
self-copy
molecule
such
as
RNA
was
exceedingly
improbable.
B
Energy
driven
networks
of
small
molecules
afford
better
odds
as
the
initiators
of
life.
So
this
is
where
this
person
has,
you
know,
issued
the
RNA
Paradigm
and
in
fact,
I
think
a
lot
of
people
have
for
one
reason
or
another,
and
so
this
is
kind
of
the
perspective
that
we're
coming
from
I
think
the
the
particular
perspective
we
were
coming
from
wasn't
really
well
stated
in
the
original
paper.
B
Inspire
extraordinary
claims-
that's
James,
Watson
reported
that
immediately
after
they
have
uncovered
the
structure
of
DNA
Francis
Crick
winged
into
the
eagle
Pub,
which
is
a
problem
where
they,
you
know
to
tell
everyone
within
hearing
that
we
had
discovered
the
secret
of
life.
So
when
they
were
working
on
this
Francis
Crick
goes
into
a
pub
and
you
know,
starts
bragging
about
discovering
the
secret
of
life.
There's
structure
and
elegant
double
helix
almost
merited
such
enthusiasm.
B
B
The
arrangement
suggested
a
mechanism
for
reproduction
subsequently
Illustrated
many
biochemistry
texts,
the
two
strands
of
DNA
reported
company
as
they
are
replicated
as
they
did
so
that
DNA
new
DNA
building
blocks
called
nucleotides
lined
up
along
their
separated
strands
and
linked
up
through
double
helices,
now
existed
in
place
of
one
each
replica
of
the
original,
so
DNA
when
it
replicates
it
divides.
The
two
strands
divide,
replicate
and
then
form
their
Abdullah
helices.
B
A
B
Yeah,
so
it's
not
yeah,
eventually
it
yeah
all
right,
so
the
Watson
Creek
structure
triggered
an
avalanche
of
discoveries
about
the
way
in
which
living
cells
function.
Today,
these
insights
also
stimulated
speculations
about
life's
Origins.
So
this
kind
of
goes
through
that
and
then
many
different
definitions
of
life
been
proposed.
Mueller's
remark
would
be
in
accord
with
what
has
been
called
the
NASA
definition
of
life.
Life
is
a
self-sustained
chemical
system
capable
of
undergoing
dirt
winning
and
evolution.
B
So
you
have
to
have
the
self-sustaining
aspect,
the
self-replicating
aspect
and
has
to
be
sustainable,
so
you
get
to
life
and
then
life
has
to
be
perpetuated
and
then
it
has
to
be
able
to
undergo
some
sort
of
darwinian
evolution.
So
this
is
where
you
have
natural
selection,
replication
mutation
and
things
that
allow
for
variation
to
be
selected
on
Richard
Dawkins,
elaborated
on
this
image
of
the
earliest
living
entity.
In
his
book,
The
Selfish
Gene
at
some
point
in
particularly
remarkable
molecules
formed
by
accident,
we
will
call
it
the
replicator.
B
It
may
not
have
been
the
biggest
or
the
most
complex
molecule
around,
but
it
had
the
extraordinary
property
of
being
able
to
create
copies
of
itself.
Dawkins
wrote
these
words
30
years
ago.
Dna
was
the
most
likely
candidate
for
the
rule,
but
we've
since
then,
we've
found
several
other
replicators,
so
one
of
these
might
have
been
RNA,
so
DNA's
replication
cannot
proceed
without
the
assistance
of
a
number
of
proteins,
and
so
you
need
to
have
proteins
to
replicate,
DNA
and
so
early
life
probably
didn't
start
that
DNA,
because
you
didn't
have
the
proteins
there.
B
Proteins
like
DNA
are
constructed
by
linking
subunits
such
as
amino
acids
to
form
a
long
chain
cells
employ
20
of
these
building
blocks
and
the
proteins
that
they
make,
affording
a
variety
of
products
capable
of
Performing
many
different
tasks.
So
proteins
are
doing
all
sorts
of
things
in
the
cell,
and
they,
you
know,
allow
for
this,
these
Criterion
of
life
for
sustaining
life,
but
you
have
to
get
proteins,
and
so
you
can't
start
with
DNA
and
then
get
to
proteins
without
proteins
already
being
there.
B
The
above
account
brings
to
mind
the
old
riddle,
which
came
first,
the
chicken
or
the
egg
DNA
holds
the
recipe
for
protein
destruction,
yet
that
information
cannot
be
retrieved
or
copied
without
the
assistance
of
proteins
themselves.
So
we
don't
know
where
proteins
are
DNA,
which
one
came
first
and
so
the
proteins
came.
First
then
DNA,
of
course,
didn't
come
first,
so
you
need
to
build
your
model
on
and
say,
like
some
sort
of
proteomic
approach,
a
possible
solution
appeared
when
an
attention
shifted
to
a
new
Champion
RNA.
B
So
RNA
is
like
DNA
in
a
sense
it
has.
You
know
it
has
the
kind
of
sequence
of
DNA
it
has
I
guess
it
has
four
it's
based
on
four
nucleotides,
but
it
plays
many
roles
in
our
cells
and
it
also
has
the
ability
to
fold
into
secondary
structures
like
proteins,
although
not
quite
as
complex
as
proteins,
but
it
can
form
different
types
of
secondary
structures
such
as
hair,
pins
and
folds,
and
other
things
that
are
important.
B
For
you
know,
information
processing,
certain
rnas
vary
information
from
DNA
to
structures
which
themselves
are
largely
built
of
other
kinds
of
RNA
that
construct
proteins.
So
RNA
sort
of
serve
as
this
link
between
dnas
and
proteins.
Rnas,
of
course,
come
from
dnas,
but
rnas
could
be
because
they're
based
on
these,
these
nucleotide
building
blocks,
they
could
be
sort
of
there
in
lieu
of
DNA
so
actually
to
earn
a
can
take
on
the
farm
about
double
helix
that
resembles
DNA
or
couldn't
be
folded
into
a
single
strand.
Much
like
a
protein.
B
B
And
then
the
hypothesis
at
Life,
begin
with
RNA
was
presented
as
likely
reality
rather
than
speculation.
Yet
the
clues
I
have
cited
only
support
the
weaker
conclusion
that
RNA
proceeded
DNA
and
proteins.
They
provide
no
information
about
the
origin
of
life
itself,
which
may
have
involved
stages
prior
to
the
RNA
world
and
which
other
living
entities
are
World
Supreme.
B
Just
the
same,
and
despite
the
difficulties
I'll
discuss
in
the
next
section,
perhaps
two-thirds
of
Scientists
Publishing
in
the
origin
of
Life
Field
still
support
the
idea
that
life
began
with
spontaneous
formation
of
RNA
a
related
self-copying
molecule.
So
this
RNA
World
basically
is
like
sort
of
resolves
that
riddle
of
what
came
first,
the
TNA
or
the
protein,
but
there
may
have
been
something
that's
set
up
the
RNA
world
in
in
turn
and
which
could
be
considered
living.
So
that's
where
this
is
coming
from
this
heading.
B
C
B
So
people
I
guess
foreclose
things
like
RNA
Clays
and
things
like
that.
So
they're
they're
different
things
like
you
know,
different
ways
that
you
can
think
about
RNA
world.
But
there's
you
know
the
other
way
to
think
about
this
is
that
there
are
other
things
that
are
come
before
RNA
that
have
to
be
in
place
to
enable
our
earning
World,
which
really
kind
of
kick-starts
the
the
darwinian
processes
that
we're
interested
in
with
life
so
yeah.
So
then
it
kind
of
talks
about.
B
Let's
see
in
a
form
of
molecular
vitalism,
some
scientists
have
presumed
that
nature
has
an
innate
tendency
to
produce
at
least
building
blocks
preferentially,
rather
than
hordes
of
other
molecules
that
can
also
be
derived
from
the
rules
of
organic
chemistry.
This
idea,
at
your
inspiration
of
a
well-known
experiment
published
in
1953
by
Stanley
Miller.
So
this
is
the
Miller
uray
experiment,
which
is
where
they
took
a
bunch
of
chemicals.
B
They
throw
them
in
a
big
react:
bioreactor
they
put
electrical
charge
across
it,
and
then
they
generated
these
different
types
of
amino
acids.
So
they
were
able
to
synthesize
amino
acids,
and
they
did
this
in
1953
and
it's
interesting
that
they
did
this
experiment
years
later
and
they
actually
found
that
it
was
understating
the
effect.
B
So,
like
you
know,
in
the
original
experiment,
they
were
able
to
produce
amino
acids
in
the
replication
which
happened
like
40
or
50
years
later,
they
were
able
to
do
find
many
more
types
of
biomolecules
that
were
produced.
So
it's
definitely
not
a
fluke,
and
you
you
have
this
aspect
of
producing
amino
acids,
amino
acids
being
the
outcome
of
rnas,
so
rnas
will
encode
for
different
amino
acids
in
a
number.
So
there
are
a
number
of
amino
acids
that
come
from.
You
know.
B
You
know
you
have
these
RNA
coding
triplets
that
produce
the
amino
acids,
and
so
you
know
there
are
different
ways
that
you
can
synthesize
these
without
having
to
go
to
the
RNA
template.
But
this
is
you
know
something
they
were
able
to
do.
They
were
able
to
actually
identify
things
in
very
old
symbol:
paleo
paleontological
assemblages,
like
the
merchants
meteorite,
which
is
one
of
the
early
or
you
know
one
of
the
earliest
forms
of
sort
of
you
know
early
evidence
of
early
life.
B
Where
you
see
this
in
the
in
the
assemblage
you
see
remains
of
this
Nature
has
been
apparently
been
generous
in
providing
a
supply
of
these
particular
building
blocks.
By
extrapolation
of
these
results,
some
writers
have
presumed
that
all
of
life
life's
building
could
be
formed
with
ease
in
Miller
type
experiments
in
our
present
meteorites,
their
extraterrestrial
bodies.
This
is
not
the
case,
and
so
this
kind
of
goes
on
about
a
careful
examination
of
the
results
of
blood
scientists
who
conducted
the
work
to
a
different
conclusion.
B
Inanimate
Nature
has
a
bias
towards
the
formation
of
molecules
made
of
fewer
rather
than
greater
numbers
of
carbon
atoms.
That
shows
no
partiality
in
favor
of
creating
the
building
blocks
of
our
own
kind
of
life.
So
this
just
kind
of
goes.
You
know
through
this
aspect
of
early
life,
this
is
pre-rna
world.
D
B
A
A
B
A
C
B
B
A
A
B
B
Okay,
well
I
guess:
I
have
one
more
thing
before
we
go.
There's
this
interesting
resource
on
bioengineering
that
I
found
I
thought
it
would
be
interesting
for
people
if
you're
interested
in
by
understanding
bioengineering
or
having
a
reference
is
the
beginner's
guide
to
mammalian
genetic
design.
So
there's
been
a
lot
of
bio,
you
know
like
bioengineering
and
bacteria
and
in
bacterial
genomes,
and
this
is
kind
of
giving
us
sort
of
a
window
into
how
to
use
the
new
technologies
at
crispr,
which
is
where
you
can
manipulate
genes.
B
B
You
know
so
for
a
while
they've
been
able
to
do
things
like
engineer,
E
coli
to
produce
human
insulin
or
I.
Think
I
talked
about
in
another
meeting
where
we
have
a
really
high
ability
to
manipulate
bacterial
genomes
to
find
minimal
genomes
and
then
to
boot
up
like
specialized
organisms
that
do
things
like
clean
up
the
environment
or
they
can.
You
know
we
have
medical
uses
so,
but
this
is
for
mammalian
cells.
The
mammalian
genomes,
of
course,
are
much
different
than
bacterial
genomes.
You
know
they're
not
circular.
B
They
have
a
lot
more
in
terms
of
coding,
regions
and
Regulatory
regions
and
junk
DNA,
and
things
like
that
that
you
get
that
you
have
to
sort
of
navigate
and
understand
how
to
manipulate
to
get
things
to
work.
So
this
is
a
nice.
So
Asimov
is
a
company
I,
guess
that
that
creates
tools
for
this.
They
have
a
guide
to
this
they're
building
a
full
stack
platform
for
genetic
design,
including
software
to
design
and
debug
genetic
systems,
biophysical
models
to
simulate
cell
processes
and
tools.
B
The
engineer
living
systems
and
measure
molecules
so
they're
doing
this
sort
of
full
stack
approach.
They're,
not
just
designing
this.
You
know
guidelines
for
using
crispr.
You
know
you'll
need,
like
some
sort
of
AI,
probably
to
help
you
find
the
optimal
sort
of
way
to
do
this,
because
you're
dealing
with
pretty
large
a
lot
of
times,
you're
dealing
with
pretty
large
genomes
and
know
exactly
where
to
to
make
these
inserts.
B
You
know
it's
an
important,
but
also
you
have
these
biophysical
models
that
we
need
to
simulate
cell
processes
so
that
the
cell
processes,
you
know,
we
know
how
what
the
outcome
of
the
genetic
manipulation
is
and
then
other
types
of
tools.
So
our
goal
is
to
make
biology
a
true
engineering
discipline.
We
want
biological
engineering
to
be
robust
or
reliable
and
as
accessible
is
electrical
software.
Mechanical
engineering
so
that'll
be
interesting.
B
There's
this
project
called
igem,
which
is
where
high
school
students
will
go
and
do
these
experiments,
biological
engineering
experiments
and
it's
a
competition
that
they
have
yearly.
So
this
is
kind
of
designed
for
igem,
but
this
is
useful
if
you're
interested
in
just
getting
sort
of
an
introduction
to
some
of
this
to
this
world.
B
So
you
know
this
is
where
you
want
to
be
able
to
build
genetic
parts
and
build
like
I,
guess
the
analogy
to
machines
and
so
to
understand
some
of
the
things
that
happen
in
when
you're
in
the
process
of
doing
biological
engineering
like,
for
example,
plasmids
that
don't
work
or
break
cells
that
die
for
unknown
reasons.
You
know
it's
important
to
kind
of
have
a
way
to
work
your
way
through
those
things
and
debug
them.
So
this
is
this
kind
of
gives.
You
I
think
a
good
guide
to
like
different
cell
types.
B
You
might
use
so
like
they're
different
model
cell
types,
like
excels,
even
embryantic,
kidney
cells,
Chinese
hamster
ovary,
Cho
cells
and
Vero
cell
lines
used
to
produce
viral
vectors,
therapeutic
antibodies
and
vaccine
components.
So
these
are
the
you
know:
they're
different
types
of
cells
that
they
are
kind
of.
Favorites
of
so
we
were
reprogramming
and
some
of
the
other
bioengineering
approaches
and
they
kind
of
walk.
You
through
examples
for
those
cell
lines
says.
A
B
Of
goes
through
like
why?
What
is
genetic
design,
and
why
should
you
care?
You
know
it
tells
you
that,
basically,
you
know,
cellular
functions
arise
from
the
expression
of
their
genes.
So
a
lot
of
these
Technologies
are
used
to
control
their
G
control,
the
expression
of
genes
which
allows
cells
to
do
things.
So
that's
what
you're
trying
to
do
we're
trying
to
be
able
to
create
altered
living
cells
in
a
way,
that's
both
reproducible
and
predictable.
B
So
you
need
design
guidance
to
make
that
happen,
and
so
it's
really
kind
of
good
I,
don't
know
like
I
mean
it's
pretty
long.
Let's
see
this
just
kind
of
goes
through
a
very
general
introduction
to
this.
So
This
talks
about
RNA,
polymerase
II,
which
shows
you
the
production
of
proteins
from
DNA,
and
it
just
shows
you
the
steps
and
how
this
works.
B
So,
if
you're
engineering,
a
cell,
you
know
you
need
to
know
kind
of
the
structure
of
the
Gene
and
the
regulatory
regions,
but
also
the
structure
of
the
MRNA
and
then
the
protein
through
transcription
through
these
post-transcriptional
modifications,
translation
and
then
post-translational
modifications
which
aren't
really
shown
here.
But
that's
another
thing:
you
step
you
get
on
the
way
to
proteins,
so
there
are
a
lot
of
different
steps.
You
need
to
consider
not
just
manipulating
the
DNA.
B
B
So
it's
a
nice
introduction,
I
think
if
you're
interested
in
this
area,
even
if
you're
not-
and
you
just
want
to
know
more
about
it-
I
think
this
is
a
good
introduction.
So
they
talk
about
things
like
coding,
sequences,
they
really
kind
of
break
down
cell
biology
or
molecular
biology
into
this
sort
of
set
of
design
principles
almost
which
I
kind
of
like,
because
it's
like
the
the
molecular
biology
would
return
very
opaque
with
respect
to
like
some
of
these
things,
so
you're
you're
doing
you're
trying
to
understand
all
these
things.
B
You
generally
learn
it
by.
You
know
using
flash
cards
and
you
learn
all
these
facts.
And
then
you
know,
if
you
go
into
the
lab,
you
have
to
do
an
experiment.
You
have
to
figure
out,
like
you
know
these
parts,
how
do
they
work,
and
so
a
lot
of
that
is
this
trial
and
error,
but
having
a
good
guide
to
this
in
terms
of
design,
principles,
I
think
is
good.
A
B
A
B
B
A
B
C
C
B
C
B
B
B
C
If
I
can't
my
friend,
Dr
Sharif
will
have
a
fit.
Oh
yeah,
oh
well,
there's
lots
to
explore
with
with
the
model
and,
like
I
said,
there's
a
whole
entire
book
written
about
this
that
this
type
of
structure.
So
it's
good
it's
about
metamaterials.
Actually,
thank
you.
Would
you
like
them
title
of
the
book,
yeah
that'd.
C
C
C
You
can
a
3D
print,
very
interesting
materials
that
actually,
when
you
compress
them,
they
shrink
and
when
you
pull
on
them,
they
enlarge
like
they're
like
they're.
Poissons
ratio
is
opposite
to
what
you
think
it
should
be:
yeah,
they're,
they're,
intriguing
and
they're.
Also,
if
you
do
them
with
electromagnetic
fields,
you
can
get
cloaking
devices.
C
B
B
C
B
B
So
these
are
going
to
be
things
that
you
know
you'll
be
attending
this
meeting
and
then
probably
some
other
activities
alongside
that,
I'll
provide
some
readings
and
I
encourage
people
to
explore
the
open
worm
Community,
but
also
the
the
we
have
another
community
that
you
might
be
interested
in.
Exploring
the
diva
worm
group,
of
course,
has
done
a
lot
of
historical
work,
so
I
encourage
you
to
explore
that
as
well.
So
thanks
everyone
for
attending
and
talk
to
you
next
week.
C
B
B
B
A
new
study
traces
maturation
of
an
entire
circuit
in
some
cellular
detail
revealing
continuous
functionality
undergraduate
structural
change,
so
this
is
has
to
do
with.
This
is
more
generally
the
case
with
the
connectoms,
but
in
this
case
they're
going
to
focus
on
coal
against
connectomes,
so
they
kind
of
go
through
this
anecdote
here
at
the
beginning,
a
common
characteristic
of
complex
systems
is
that
they
stare
out
relatively
simple.
Therefore,
any
additions
or
deletions
are
constrained
by
the
previous
instantiation
of
the
system
and
required
functionality
of
the
output.
B
So
this
is
basically
describing
the
origins
of
a
connectome,
and
so
they
start
out
simple
and
then
they
become
more
complex.
A
major
open
question
in
complexity.
Science
is:
how
evolve
in
your
developing
systems
to
maintain
functionality
in
the
face
of
change,
in
other
words,
it's
more
apparent
than
in
development
and
lifelong
changes
of
narrow
circuits,
as
they
incorporate
omit
and
reappropriate
neurons
and
connections.
B
So
there's
this
paper
in
current
biology
by
Mulcahy
at
all
where
they
approach
this
question,
one
of
the
most
thoroughly
described
nervous
systems,
that
of
the
nematodes
elegans
as
its
juvenile
Locomotion
circuit,
develops
into
its
mature
configuration.
So
this
is
during
the
larval
stages,
where
the
C
elegans
is
behaving,
but
during
its
juvenile
stage
it
goes
through
a
significant
amount
of
plasticity.
With
respect
to
synaptic
connectivity,
and
so
the
question
is,
is
how
does
that
movement
circuit
change
in
larval
development?
B
So
at
hatching,
the
C
elegans
Locomotion
circuit
includes
only
22
neurons
of
three
classes:
two
cholinergic
classes
that
innervate
dorsal
muscle
in
one
Gava,
urgent
class
that
innervates
ventral
muscle
so
they're
starting
at
hatching.
So
this
is
right
when
the
C
Elegance
hatches
onto
the
egg
and
then
goes
into
the
stage
larval
stage
L1,
and
so
we
can
look
at
the
neurons
that
exist
in
The
Locomotion
circuit.
B
At
that
point,
53
more
neurons
of
five
new
neuronal
classes
are
integrated
post-embryonically.
So
this
is
something
that
happens
where
more
neurons
are
added
to
that
circuit
and
specified
for
movement.
The
more
than
tripling
of
the
circuit
raises
two
over
urchin
questions.
How
does
the
asymmetric
juvenile
circuit
generate
the
alternating
muscle
pattern
needed
to
drive
undulatory
locomotion?
B
So
that's
the
first
question.
The
second
is:
how
is
its
catastrophic
failure
avoided
during
the
transition
from
the
juvenile
to
the
mature
circuit?
The
first
question
was
addressed
by
another
paper,
also
in
this
issue
of
current
biology.
This
is
Lou
ahamed
at
all.
So
these
are
the
authors.
This
demonstrate
that
during
rhythmic
undulations
in
either
forward
or
backward
Direction.
So
when
the
worm
wriggles
back
and
forth
its
characteristic
movement,
is
you
can
look
at
any
video
of
the
worm
and
see
this
characteristic
movement?
B
B
So
this
is
a
sort
of
a
regulatory
thing
that
they're
observing
as
to
the
second
question
until
now,
neuroscientists
believe
the
Swedish
model,
which
is
the
new
poster
embryonic
infrastructure,
was
put
in
place
within
hours
of
animal
hatching
in
the
first
larval
stage.
So
basically,
a
lot
of
this
infrastructure
is
already
in
place
upon
hatch,
and
so
this
is
the
Swedish
model,
which
is
the
I
guess
old
model
and
we're
going
to
be
proposing
a
new
model.
B
This
is
reference
10,
so
we'll
get
to
the
references
in
a
minute,
but
for
right
now
keep
that
in
mind's
reference
10.
and
the
switch
was
then
assumed
to
be
orchestrated
during
a
sleep
state
that
occurs
between
the
first
and
second
larval
stages,
so
between
this
L1
and
L2.
But
C
elegans
never
cease
to
surprise
us,
which
is
that
first
paper
we
talked
about,
shows
that
the
remodeling
continues
after
the
Sleep
state.
So
the
circuit
must
maintain
functionality
in
the
awake
state
as
the
circuit
matures
along
the
body.
B
B
So
C
elegans
allows
us
to
do
a
lot
of
these
sorts
of
things.
To
look
at
these
circuits
in
very
Exquisite
detail
with
invariant
and
fully
mapped
cell
lineages
anatomical
changes
can
be
traced
developmentally
across
different
animals
at
the
subcellular
and
even
synaptic
resolution.
This
is
true
of
the
Mulcahy
paper.
They
use
two
state-of-the-art
surreal,
electron
microscopy
methods
to
reconstruct
the
neuronomorphology
and
neural
apposition
of
The
Locomotion
circuit
at
six
different
developmental
stages
before
during
and
immediately
after
the
circuit
is
reorganized
to
the
mature
configuration.
B
This
confirmed
what
we
already
knew
about
this
three-stage
process.
Neurogenesis
occurs
early
on
in
a
neat
staggered
fashion
in
time
along
the
body.
Next
embryonic
neurons
and
their
synapses
undergo
significant
remodeling
neurons
of
the
gabaergic
embryonic
class,
dramatically:
reverse
their
external
dendritic
polarity
and
switch
from
innervating
ventral
to
Dorsal
muscles
remodeling
follows
a
similar
spatio
temporal
order
from
head
to
tail.
Finally,
neurons
growing
form
synapses
with
the
rest
of
the
circuit
muscles.
So,
what's
new,
they
describe
a
transition
in
fine
detail
and
extract
several
rules
of
organization.
B
So
there
are
several
sort
of
overarching
principles
here.
First,
they
describe
the
secession
of
Developmental
processes,
neurogenesis
neurotic
growth,
synaptic
formation
synaptic
pruning
in
this
order,
as
they
occur,
sequentially
and
iteratively
along
the
anterior
posterior
axis
from
head
to
tail.
So
you
have
neurogenesis,
which
is
the
birth
of
neurons
neurite
growth,
which
is
the
growth
of
the
outgrowths
of
the
neurons
synaptic
such
as
axons
and
synapses
synaptic
formation
themselves
and
then
synaptic
pruning.
B
B
Second,
the
process
of
neuron
growth
is
hallmarked
by
a
growth
cone,
wrapping
around
future
synaptic
Partners
other
neurons
and
non-normal
tissue
leaving
neural
tracks
in
their
place,
reminiscent
of
intercellular
communication
or
envelope
and
other
systems.
So
this
is
where
we
set
up
this
communication,
these
communication
Pathways
in
the
nervous
system.
B
They
have
dorsal
de
ventral.
This
is
the
top
to
the
bottom.
This
is
dorsal
muscle
and
inhibitory,
neurons
rhythmically
excited
by
cholinergic
neurons.
So
this
dorsal
ventral
organization
here
is
where
it's
producing
these
rhythmic
movements,
it's
producing
it
in.
You
know
what
the
dorsal
Direction
versus
the
ventral
Direction
and
it's
moving
back
and
forth
along
that
axis.
So
it's
moving
forward
it
undulates
in
One,
Direction
backwards
and
on
joints
in
another
Direction.
B
Then
there's
this
functioning
transition,
where
neurogenesis
and
incorporation
of
53,
neurons
of
five
classes
happens
and
plus
the
remodeling
of
these
existing
22
neurons
of
three
classes.
So
you
not
only
get
new
neurons
added
into
this
network,
but
remodeling
of
the
old
Network
and
then
the
mature
circuit
75,
neurons
of
eight
classes.
So
now
you
get
this
mature
State.
You
have
dorsal
muscles
excited
inhibited
by
dorsal
neuron
classes.
Ventral
muscle
is
excited
and
inhibited
by
ventral
neuron
classes,
so
you're
getting
like
specific
neurons,
specific
parts
of
the
anatomy
dorsal
to
ventral.
B
You
know
it
sort
of
persists
and
it
doesn't
have
to
shut
down
every
wire
and
open
back
up
again,
like
you
know,
like
a
carnival
ride
or
something
so
this
is
you
know
what
we
want
to
do
is
we
want
to
be
able
to
develop
mathematical
and
computational
models
that
allow
for
us
to
describe
these
kind
of
systems?
We've
done
a
little
bit
of
work
on
this
with
respect
to
the
embryo.
Certainly,
there
are
a
lot
of
network
models
that
could
be
used
for
this.
B
There
are
a
lot
of
dynamical
network
models
that
could
be
used
for
this.
So
we'll
probably
revisit
this
again
talking
about
some
of
the
things,
maybe
that
you
know
there
are
other
additional
papers
that
we
can
find
and
draw
from.
If
we
want
to
do
some
sort
of
research
in
that
area,
let's
talk,
but
why
don't
we
look
at
some
of
these
citations?
Now
we
have
a
couple
of
citations
here
that
we
wanted
to
focus
on
so
number
10.
B
B
There's
this
other
paper
number
nine
extra
synaptic
signaling
enables
an
asymmetric
juvenile
motor
circuit
to
produce
symmetric
undulation.
So
this
is
a
paper
of
the
problem.
This
I
guess
this
issue
of
current
biology
says
Daniel
Whitfield
on
it
and
okay.
He
and
some
other
people
on
it.
So
this
is
sort
of
this.
This
was
actually
a
paper
where
Daniel
Whitfield
released
a
paper
in
science.
B
This
is
the
this
is
actually
the
other
paper
by
Daniel,
Whitfield
and
I.
Think
there's
more
keys
on
this
as
well.
So
this
is
connectoms
across
development
reveal
principles
of
brain
maturation.
That's
the
paper
I
was
referring
to.
Then
we
have
these
minimum.
This
paper
minimal
model
of
C
elegans
forward
Locomotion
the
larval
L1
circuit.
This
is
by
boil
on
netacone
and
we
have
I,
don't
think
we
have
any
other
papers
pick
out
specifically
here.
B
Let's
see,
let's
see,
probably
that's
enough
for
now.
There's
another
paper
that
I
wanted
to
get
to
here,
and
this
focuses
on
a
specific
stage
of
learnable
to
build.
So
this
is
a
mind
of
a
tower
comparative
connectomics
reveals
developmental
plasticity.
So
this
is
from
a
Korean
group
where
they're
looking
at
the
dower
stage
and
the
dower
stage
is
sort
of
a
polyphemism.
B
So
it
starts
at
L2,
which
is
a
Marvel
stage
two
and
goes
to
L4
and
the
reason
the
dour
stage
exists
is
in
cases
where
the
C
elegans
experiences
extreme
stresses
things
like
droughts
or
lack
of
food,
and
so
they
go
into
this
larval
stage
the
tower
stage,
which
is
a
polyphenism.
That's
that
triggers
a
number
of
neural
protective
changes,
changes
in
the
in
the
cuticle
changes
in
the
neuronal
connectivity
and
basically
it
goes
into
quiescent
state.
So
it's
not
completely
quiescent,
but
it's
a
semi
quiescent
state.
B
B
So
the
abstract
reads:
a
fundamental
question
in
neurodevelopmental
biology
is
how
flexibly
the
nervous
system
can
change
during
development.
To
address
this
question
of
Developmental
plasticity,
we
analyze
the
connectome
of
Tower
and
alternative
developmental
stage
of
nematodes
with
physiological
and
behavioral
characteristics,
remarkably
distinct
from
other
developmental
stages.
B
Their
specific
behaviors,
you
can
observe
that
are
distinct
from
other
C
elegans
larval
stage
of
similar
wearable
stages,
combining
the
connectome
data
and
optogenetic
experiments
were
enough
to
reveal
our
specific
neural
connections
for
the
Delaware
specific
behavior.
So
there
are
neural
connections
that
are
downward
specific
that
translate
to
these
behaviors
graph.
Theoretical
analysis
shows
higher
clustering
or
motor
neurons
and
more
feedback
connections
from
motor
to
Sensory
neurons
in
the
tower
connected.
So
they
did
a
graph
theoretical
analysis
of
the
connectome,
which
is
generally
what
you
do
but
to
see
where
these
connections
these
feedback.
B
In
this
case,
feedback
connections
and
clustering
of
motor
neurons,
occur
so
their
changes
that
basically
allow
for
the
motor
neurons
to
become
more
connected
around
hubs.
I
guess
and
then
have
more
feedback
connections,
remote
or
just
Sensory
neurons,
so
you're,
getting
feedback
from
sensory
motor
neurons
in
the
connectome,
suggesting
that
the
dower
connect
Dome
allows
a
quick
response
to
an
ever-changing
environment.
So
it
allows
it
to
pick
up
more
environmental
information
and
act
on
and
work
weekly.
B
We
suggest
that
the
nervous
system
and
the
nematode,
which
can
be
extended
knit
to
animals
in
general,
have
evolved
to
make
obtain
the
ability
to
respond
to
harsh
environments,
a
reversively
developing
a
connectome
quantitatively
and
qualitatively
differentiated
from
other
developmental
stages.
So
this
is
a
specific
stage
in
the
connect
in
the
sort
of.
B
So
we
have,
let's
see
if
we
have
any
images
here,
sometimes
they're
at
the
end
of
the
paper,
so
here's
some
images
just
to
underscore
what's
going
on.
So
this
is
the
first.
This
is
a
map
of
what
tower
looks
like
in
terms
of
the
life
cycle.
So
you
have
this
embryo,
the
L1
stage
in
the
L2
stage,
so
the
tower
Folly
phenism
sort
of
happens
red
between
L1
and
L2.
B
And
if
you
go
down
this
pathway,
you
go
through
L2
l3804,
which
are
just
you
know
where
the
the
word
grows
and
it
starts.
You
know
you
get
the
changes
that
we
saw
in
the
last
paper
in
the
dollar
stage.
It
goes
from
L1
and
it
goes
through
the
stage
where
there
are
a
bunch
of
different
adaptations,
both
phenotypic
and
conectomic,
and
then
you
end
up
at
L4.
So
you
know
it
emerges
in
L4
and
then
it
becomes
an
adult.
B
B
B
You
have
this
gfp
stain
that
shows
in
you
know,
in
the
adult,
in
the
L3
stage
in
the
adult
stage
these
are
without
you
know.
These
are
not
our.
This
is
a
dower
stage
where
you
get
this,
this
gfp,
which
is
different
from
the
adult
in
the
L3,
and
then
you
have
the
post-door
adults.
So
these
these
changes
actually
may
be
effect,
Downs
down
the
nerve
Court
here.
B
So
you
see
a
little
bit
of
a
persistence
at
that
adaptation
and
post-door
adults,
but
not
as
much
as
you
see
during
dollar,
with
respect
to
comparing
it
to
something
like
L3,
you
see,
L3,
the
adult
and
Tower.
You
see.
These
are
different
parts
of
the
nerve,
the
nerve
ring,
and
then
you
have
L-3
dolphin
Tower.
This
is
for
urb.
This
is
for
rih,
so
these
are
specific
cells
connected
into
that
so
yeah.
So
this
is
showing
the
branches
for
urb
neurons
here
the
branches
for
RH
neurons
here.
D
B
It
just
shows
the
difference
between
the
tower
stage
and
some
of
the
other
stages.
This
actually
shows
the
completion
of
the
Delaware
connectome.
So
now
they
have
a
dower
connectum.
Basically
you
have
the
Dower
connectome
it
it's
conserved
with
a
lot
of
parts
of
the
Ember,
the
different
stages
of
the
larval
connect
Dome.
So
L1
you
have
L3
of
you
have
L1
L2
l304.
We
have
the
adults,
you
have
a
couple
of
adult
connect
domes.
B
Then
you
have
the
Dow
or
specific
connect
Dome,
where
they're
499
the
these
are
synaptic
connections,
they're
stage
conserved,
they're,
common
317
connections.
There
are
31
that
are
lost
in
the
Dover
they're
1995
total
won
the
dollar.
So
you
get
this
this
variation,
you
get
some
that
are
conserved,
Over,
All,
forms
of
adult
and
larval,
and
then
you
get
some
that
are
lost
specifically
in
the
tower
and
you
get
some
that
are
it
gained
specifically
in
the
tower.
B
So
this
look.
This
is
a
comparison
of
normal
developmental
stage
in
the
adult
versus
the
dollar
stage.
So
you
see
that
there's
some
changes
in
the
weights
of
connections,
so
you
get
some
hubs
that
are
more
densely
connected,
as
mentioned
abstract,
and
then
you
get
these
other
interior
morphological
changes
and
connectional
changes.
B
Okay,
so
that's
that
paper
now
the
third
paper
and
the
final
paper
is
a
paper
and
canalization
of
plasticity
on
the
developmental
manifold
of
C
elegans,
and
so
this
is
by
David
Jordan
and
Eric
Miska
and
they're
from
the
garden
Institute
at
Cambridge
and
so
a
lot
of
times.
We
talk
about
the
sort
of
development
and
developmental
plasticity
and
something
called
canalization
on
a
structure
called
a
an
epigenetic
landscape,
and
so
the
epigenetic
landscape
basically
has
these
channels
where
the
developmental
trajectories
unfold.
They
call
that
process
canalization.
B
So
development
follows
a
certain
trajectory
or
certain
set
of
trajectories
based
on
what's
most
likely
to
occur
in
development,
and
we
saw
that
with
some
of
these
examples
in
the
connectome,
where
you
have
changes
in
the
connect
Dome
that
can
emerge,
but
those
changes
are
limited
to
what's
already
there.
It's
also
limited
to
function,
functional
like
plausibility,
and
things
like
that.
B
So
this
this
is:
how
does
the
same
mechanism
that
Faithfully
regenerate,
complex
developmental
programs
and
spite
of
environmental
and
genetic
perturbations
also
permit
responsiveness
to
environmental
signals,
adaptation
and
genetic
Evolution?
So
that's
the
question.
So
basically,
how
do
we
get
complex?
How
do
we
get
mechanisms
that
reproduce
complex,
developmental
programs
that
need
to
be
buffered
from
genetic
perturbations
or
environmental
perturbations,
but
also
can
be
responsive
to
these
signals?
So
how
can
something
be
adaptive
but
also
protective?
B
You
don't
want
just
anything
expressed
through
development.
You
want
these
specific
plausible
phenotypes
to
be
expressed
and
persist
throughout
development
using
the
nematodes,
the
elegans.
We
explore
phenotypic
space
of
growth
and
development
and
various
genetic
and
environmental
contexts,
our
data,
our
growth,
curves
and
developmental
parameters
obtained
by
automating
microscopy.
B
So
that's
an
interesting
statement,
so
we
have
this
developmental
space
and
then
we
find
these
correlations
and
there's
both
within
and
among
variation
and
So
within
particular
context.
We
have
correlations
that
predict
things
among
different
contexts,
so
there's
a
linkage
there
further
we
find
that
the
developmental
variability
of
this
animal
can
be
captured
in
a
relatively
low
dimensional,
phenotypic
manifold,
and
so
what
they're
saying
is
that
there's
they
actually
spell
phenotypic
or
I?
Guess
phenotypic.
B
Is
different
from
the
Traditional
epigenetic
landscape
or
not
I'm,
not
really
sure
I
I
haven't
read
this
paper
in
depth
and
detail,
so
we
might
come
back
to
this
paper
later
once
we've
had
time
to
mull
over
what
what
they're
talking
about
and
then
on
this
manifold,
genetic
and
environmental
contributions
to
plasticity
can
be
deconvolved
independently.
So
one
of
the
problems
you
have
with
development
is
you
have
genetic
factors
and
then
you
have
environmental
factors,
and
those
factors
can
often
be
commingled
in
a
way
that's
hard
to
pull
them
apart.
B
So
in
in
essence,
you
don't
know
which
is
responsible
for
what
changes,
so
we
can
deconvolve
them
independently,
perhaps
on
one
of
these
manifolds
it'll.
Allow
us
to
understand
it.
So
that's
how
this
kind
of
differs
from
the
genetic
lands,
because
the
epigenetic
landscape
is
just
a
metaphor.
It
often
can
be
modeled
mathematically,
but
it
doesn't
allow
you
to
pull
out
the
genetic
and
environmental
contributions.
B
B
B
Approach
so
so
they're
interested
in
the
robustness
of
development,
which
means
that
development
is
robust
to
perturbations.
Yet
it's
also
adaptive
and
so
such
robustness
arises
at
many
spatial
and
temporal
scales.
For
example,
gene
expression
patterns
give
rise
to
reproducible
cell
differentiation,
neuromuscular
activity
generate
Locomotion
and
interactions
between
individuals
that
different
species
give
rise
to.
Surprisingly,
reproducible
ecological
Dynamics,
this
robustness
is
called
canalization
and
the
Dynamics
that
are
canalized
are
said
to
be
homo,
homologic,
so
homeoretic
and
canalization.
B
B
B
So
there's
this
concept
where
you
know
you
can
take
every
expression
and
every
type
of
face
in
the
world
human
face,
and
human
expressions
and
I'll
put
them
into
an
algorithm
and
analyze
them
and
an
output
being
the
eigenface,
which
is
a
multi-dimensional
reduction
of
all
that
variation
or
a
few
characteristic
Dimensions
or
eigen
eigenvectors.
And
then
you
have
these
eigenfaces,
which
are
the
eigenvectors
that
result.
So
this
is
this
is
basically
the
idea
you're
using
dimensionality
reduction
you're
using
a
PCA
type
approach
to
produce
eigenfaces.
B
The
openworm
group
actually
worked
on
a
paper
a
long
time
ago,
where
they
produced
these
eigenworms
or
these
eigen
poses
for
C
elegans
movements,
so
the
C
elegans
is
moving
around
and
they
were
able
to
take
images
of
these
and
characterize
them
as
these
eigen
eigenworms
or
eigen
I
care
or
the
term
that
was
used.
But
it's
the
same
principle
there
too,
and
this
is
an
example
of
eigenfaces,
though
the
eigenfaces
are
modes
of
the
system.
Varying
the
weights
of
each
mode
can
generate
many
diverse
faces.
B
The
Morpho
space
is
every
possible
combination
of
that
shape
and
that
phenotype
and
but
you
only
see,
select
phenotypes
in
that
space
that
are
actually
manifest
in
in
organisms,
whether
those
are
the
highest
Fitness
phenotypes
or
whether
those
are
just
the
phenotypes
that
result
from
evolutionary
constraints.
We
don't
really
know,
but
we
do
know
that
these
low
dimensional
spaces
can
describe
these
limited
number
of
states
which
can
of
course,
exhibit
variation
but
they're.
B
B
So,
additionally,
well
the
dimensionality
of
dynamic
time
varying
and
responsive
phenotypes
is
more
difficult
to
Define
rigorously.
It
has
been
shown
in
some
cases
to
be
low
dimensional.
So
this
is
like
the
crawling
behavior
of
C
elegans,
like
we
just
discussed
and
the
neuronal
Dynamics
that
underlie
it,
assuming
behavior
of
both
eukaryotic
and
prokaryotic
single
cell
organisms
and
the
transcriptional
trajectories
of
cells
during
feed
determination.
B
So
they're
all
these
examples
we
love
literature,
concentration
of
Dimension
may
be
an
intrinsic
property
of
systems,
but
a
robustness
and
control
the
macroscopic
observables
are
required
and
may
arise
from
constraints
imposed
by
a
steady
state,
growth
or
tradeoffs
between
phenotypic
architectures
suited
for
different
tasks.
So
it
might
be
that
you
have
these
these
discrete
States
in
life
that
are
sort
of
optimal
or
that
emerge
as
the
result
of
you
know,
exploration
of
that
space.
B
B
B
So
here
we
propose
the
process
of
canalization
may
be
superseded
by
a
more
General
process
of
phenotypic
concentration
of
Dimension.
So
this
is
where
we
have
this
concentration
of
Dimension.
So
all
the
dimensions
of
the
phenotype
can
be
reduced
down
to
only
a
few.
So
phenotypic
variation
can
be
explained
by
only
a
few
factors,
or
is
it
something
to
think
that
it's
a
very
complex
process
that
they're,
you
know
thousands
of
genes,
maybe
in
an
organism
there's
inner?
B
There
are
interactions
between
genes
and
environment,
and
then
you
get
this
very
complex
phenotype,
which
you
would
ostensibly
think
would
have
a
lot
of
Dimensions
to
it.
In
fact,
it
may
be
that
you
have
very
few
Dimensions
to
worry
about,
so
that
is,
genetic
networks
evolve
such
that
most
variations
will
be
projected
onto
a
low
dimensional
manifold
and
that
this
in
turn
provides
both
canalization
and
phenotypic
plasticity.
B
D
B
Your
favorite
is
so
we
can
actually
look
to
the
developmental
program
of
drosophila,
which
is
the
fruit
fly,
and
in
this
paper
they
use
the
landmark
framework
for
metrics,
which
means
they
just
use
like
Mark.
You
know
Morpho
morphological,
markers
they're,
probably
using
some
sort
of
imaging
technology
to
pick
out
different
things.
It's
Landmark
free,
so
they're
just
kind
of
characterizing
the
shape
and
they
were
doing
that
they
were
able
to
unco
uncover
a
dominant
amount
of
variation
that
was
not
Apparent
from
traditional
Landmark
based
methods.
B
C
B
B
This
is
time
from
hatching
length
and
microns.
This
kind
of
gives
a.
So
that's
what
they
say
here.
C
is
the
full
developmental
time
course
of
an
animal
from
hatching
to
adulthood.
The
Blue
Points
show
calculated
length
over
time
with
10
specific
points.
High
weighted
V
shows
the
length
measurements
in
the
computed
logistic
best
fit
curves
from
three
examples
of
the
time
course.
The
parameters
of
each
logistic
fit
are
shown
to
the
left
which
each
color
corresponding
with
relevant
data
and
curve.
B
B
So,
for
each
growth
curve
we
have
an
independent
measurement
of
the
reproductive
developmental
time
tdev
and
the
Animals
linked
up
this
time.
L
t
laid
L
Dev
equals
L
Dev,
so
T
Dev
and
L
Dev,
basically
developmental
time
and
length
at
the
time.
We
can
use
these
additional
parameters
to
rescale
each
growth
curve.
So
we
want
to
characterize
this
sort
of
logistic
growth.
B
B
This
figure
actually
shows
a
relationship
between
genetic
architecture,
dimensionality
and
the
correlation
structure
of
traits.
So
here
you
have
this
Gene
regulatory
Network.
Where
you
have
this
regulatory
element
that
turns
on
a
gene,
we
can
use
that
to
mount.
You
can
use
those
models
to
predict
growth
of
the
worm
here.
So
we
have
these
different
low,
dimensional
dimensions
and
we
can
actually
characterize
the
gene
expression
Dynamics
in
the
form
of
this
function.
B
So
you
get
this
within
and
between
population
variation,
so
you're
able
to
map
between
the
genotype,
the
gene,
expression,
networks,
the
gene
regulatory
Networks,
the
phenotype
and
then
this
manifold.
And
then
you
can
characterize
this
in
terms
of
this
this
within
in
between
sample
population
sample.
So.
B
This
basically
shows
this
approach.
You
have
these
developmental
parameters
that
grow
weight
with
each
popular
with
each
population.
Measured.
The
scatter
plot
shows
z-scores
of
these
parameters
with
respect
to
the
conditions
and
mean
and
variance
so
these
are
the
z-scores
here
so
you're
able
to
show
the
mean
and
variance
around
these
functions.
B
B
So
I
guess
this
was
successful
in
the
take-home
message.
Here
is
basically
that
we
have.
We
can
characterize
a
sort
of
canalization.
We
can
characterize
analyzation
with
respect
to
developmental
flexibility,
which
is
this
phenotypic
variation.
We
can.
We
can
reduce
this
phenotype,
which
is
thought
to
be
a
very
high
dimensional
space,
a
very
complex
set
of
possible
possibilities
to
the
slow
dimensional
manifold
that
is
predictive.
That's
predicted
from
it's
like
gene
expression
to
the
phenotype.
So
this
is
a
really
interesting
paper.
B
It
is
like
a
a
different
view
on
things
with
respect
to
the
epigenetic
landscape,
and
so
we
can
connect
this
to
statistical
mechanics
and
some
of
the
predictive
value
of
that,
and
so
what
they
call
the
particular
approach.
They
call
Core
screening
but,
of
course,
bringing
a
mineral,
many
interacting
degrees
of
freedom.
We
can
boil
down
to
a
few
dominant
modes
that
can
be
formulated
precisely
in
terms
of
projection
operators
and
that's
kind
of
what
they're
getting
at
here.