►
From YouTube: DevoWorm #34: Project updates (Cancer Imaging, DevoLearn), C. elegans synapse and neuronal diversity
Description
Project updates for Cancer Imaging, what's next for DevoLearn. Neuromatch 5 preview. Components of C. elegans Connectome (synapses and neuronal diversity). Connections to tensegrity and mitochondrial energy balance. Attendees: Bradly Alicea, Richard Gordon, Susan Crawford-Young, Jesse Parent, and Harikrishna Pillai
B
B
B
Yeah
and
I
think
the
first
step.
I
I
know
you
guys
are
working
on
the.
What
do
you
call
it?
The
noise,
the
noise
problem,
but
I
forgot
the
name
of
it,
okay,
but
the
first
step
really
has
to
be
a
cad
design
for
the
instrument,
which
then
gives
everybody
the
constraints
within
which
the
software
should
work.
B
Okay,
and
it
will
also
visualize
the
problem,
much
better
for
everybody.
If
we
all
agree
on
if
we
agree
on
the
visualization
I've
made
rough
sketches
of
it,
but
it
needs
a
decent
CAD
three-dimensional,
CAD
program
and
yeah,
and
so
there
I
started
that
with
my
dick
he's
downloaded
it
onto
his
computer.
The
program
unfortunately
doesn't
work
on
my
computers,
but.
A
A
B
Okay,
so
I'd
say
the
first
step
is
the
cad,
and
the
second
step
is
the
the
noise
brought
for
me
representing
the
noise,
because
then
it'll
be
within
the
context
of
the
of
the
standard
side,
and
we
actually
have
a
very
sophisticated
design,
which
I
think
is
practical.
But
this
is
one
reason
to
put
it
in
the
cad
to
to
see
I
don't
know
if
anybody
has
any
real
CAD
experience
but
or
machine
construction
experience,
but
we
have
to
make
a
design
that
can
actually
be
built.
B
Yeah,
yeah,
okay,
we're
not
going
to
build
it,
because
that
that
building
it
would
cost
hundreds
of
thousands
of
dollars.
Okay,
great,
but
I
figure.
The
total
cost
should
not
exceed
that
of
a
standard
mammography
unit,
which
is
typically
the
the
high-end
units
are
about
275
000.,
which
is
modest
compared,
for
example,
to
computer
tomography,
which
starts
typically
at
a
million
or
MRI,
which
is
typically
three
million
okay
yeah.
So
we're
not
we're
not
designing
an
instrument
that
is
out
of
the
range
of
of
current
expectations.
B
B
C
A
C
C
A
C
A
C
B
B
C
A
B
A
B
B
A
Yeah
yeah
I,
don't
know
yeah
I,
don't
know
they
had
some
issues
with
the
dashboard.
When
I
was
going
in
to
do
the
evaluations
so
I
don't
know,
maybe
there
may
be
a
bug
in
there
too.
So
if
you
don't
see
it
yeah,
let
someone
know
I,
don't
know,
I'll
ask
others
who
have
spots
yeah
because
it
was
yeah.
I
was
doing
some
weird
stuff
when
I
was
in
there
and
they
kept
emailing
every
everyone
kept
dealing.
Everyone
else
about
it
and
it
was
yeah.
A
A
A
And
this
is
where
we
keep
the
devil
and
software
here.
So
this
is
the
the
software
that
was
developed
in
like
20
19
2021,
and
then
we
have
evil
learn
is
the
Repository.
Does
that
mean
software?
Then
we
have
the
data
science,
demos,
which
actually
Elan
made
a
push
to
a
couple
weeks
ago.
So
there
are
a
number
of
data
science
Demos
in
here
and
if
you're
interested
in
doing
a
demo
of
something
you
know
you
can
upload
something
to
this
space.
A
And
you
know
we
have
a
number
of
different
things
like
Gans
Networks,
other
tutorials
centroid
extraction.
So
there
are
a
lot
of
different
tutorials
that
people
have
prepared.
Dc
Gan
I
mean
they're.
This
is
a
little
bit
I.
Don't
know
why
this
is
out
here,
but
this
is
yeah
my
Knox
tutorial,
so
there
are
a
bunch
of
different
tutorials.
A
But
this
is
the
repo
and
this
was
just
cloned
from
what
Gia
hung
had
and
he
was
working
for
repo
in
his
GitHub
repository
and
then
he's
got.
You
know,
we've
got
all
these
different
things
here.
So
I
guess
you
start
with
the
readme,
and
this
just
tells
us
who's
been
participating
and
the
different
contributions
so
far
and
I
think
this
is
something
you
have
to
download
and
run
on
your
machine
and
then
it'll.
A
It's
based
on
an
original
approach
to
segmenting
cells
and
then
building
these
graph
embeddings,
and
so
this
should
give
us
some
I
I.
You
know
I
mean
it
gives
us
a
starting
point:
anyways
we
were
going
to
submit
to
a
conference,
but
we
didn't
think
that
we
were
ready
for
it.
Yet
in
terms
of
the
work
I
asked
him.
If
he
wanted
to
go
for
it,
he
said
no,
but
that's
okay,
I
mean
that's,
probably
pretty
competitive,
so
I
didn't
want
to
force
it
on
him.
A
You
know
because
it's
a
commitment
to
get
the
paper
ready
and
everything
so,
but
we're
going
to
be
working
on
that
over
the
fall,
probably
refining,
that
a
little
bit
and
getting
it
squared
away.
I
haven't
figured
out
how
to
present
the
digital
microsphere
stuff,
yet
I
might
actually
incorporate
that
into
the
Devo
evil,
learn
organization
or
I
might
do
it
a
different
way.
I
haven't
decided
yet,
but
anyways.
That's
what
we
have
that's
we're
following
up
from
on
the
on
the
different
projects
that
we're
doing
this
summer.
A
Its
scope
includes
machine
learning,
work
that
has
an
explicit
biological
link
and
so
I
actually
presented
on
the
diatoms
when
you're
on
behalf
of
the
group
visit,
neuromatch,
3
I
believe,
which
was
I
think
at
least
a
year
ago.
It
could
have
been
two
years
ago
now,
but
you
know
this
is
neuromatch
5
coming
up
this
week,
so
this
is
September
2022.
This
is
neuromatch
five,
let's
see
if
I
go
to
the
agenda,
the
agenda
is
it's
it's
an
interesting
conference
because
it's
all
sort
of
asynchronous
what
they
do.
A
Is
they
have
the
they
have
a
main
agenda
here
with
different
sessions?
The
different
sessions
are
hosted
by
people.
So
you
have,
you
know
people
invited
speakers,
you
have
panels,
you
have
Keynotes
and
the
keynote
in
everything
is
recorded
on
crowdcast,
which
is
a
platform
that
crowdcast.io,
which
is
a
platform
that
live
streams,
video
and
then
saves
it.
It's
kind
of
like
any
live
stream
that
you
might
do.
You
can
do
this
on
YouTube,
but
it
you
know
it
the
thing
about
crowdcast
and
you
have
to
pay
for
a
crowdcast.
A
But
the
thing
is
you
can
host
like
a
live
stream
where
people
can
come
in
and
and
give
comments
during
the
live
stream,
and
then
it
manages
that
pretty
well
and
then
automatically
saves
it
as
an
archived
video.
So
all
this
will
be
archived
but
it'll
be
going
on
in
real
time,
and
so
they
have
a
session
for
like
Europeans
and
people
in
Asia.
It
starts
at
2
A.M,
my
time,
which
is
North,
America,
Central
Time,
and
then
that
goes
through
to
about
6
A.M.
A
So
that's
obviously
I'm
not
going
to
catch
that
live,
but
other
people
will
and
then
they'll
start
up
again
at
10
A.M
by
time,
going
to
1
30
p.m.
So
that's
for
North
America,
you
know
that's
pretty
friendly
for
North
America,
so
yeah,
that's
and
then
it
goes
on
for
two
days
and
they
have
these
diff
these
panels.
These
invited
Keynotes
and
things
like
that.
Then
they
have
the
they
have
these
flash
talks,
and
this
is
where
we
have
actually
my
other
group
that
I
work
with
has
four
submissions
here.
A
A
So
this
is,
you
know
where
you
these
are
actually
totally
asynchronous.
You
click
on
it,
abstract
you
go
to.
They
usually
have
a
video
or
they
have
a
link
for
a
pre-print
or
publication.
So
you
can
actually
catch
up
on
this
asynchronously.
You
just
go
to
the
abstract
browser.
You
click
on
something,
you
look,
you
go
to
the
link
and
you
can
get
access
to
the
to
the
video
and
maybe
some
other
materials.
A
So
that's
a
nice
model
for
comp
virtual
conferences.
They've
been
working
on
it
for
a
couple
of
sessions.
Now
so
there's
an
iterative
design
process
where
they
I
think
they
started
out
doing
Zoom
rooms
and
then
getting
like
what
they
call
Zoom
bombed,
which
was
where
people
take
the
link
and
sneak
in
and
do
bad
things,
but
I
think
that's
they
figured
out
a
way
to
you
know
really
make
this
optimal
optimize.
This
conference
experience,
so
that's
going
on
this
week,
yeah,
okay,
so
that's
that's!
A
All
I
wanted
to
mention
in
terms
of
like
announcements,
they're
also
some
other
C
Elegance
papers
that
have
come
out
recently.
I
think
I
talked
about
a
couple
last
week.
One
was
reproduction
and
the
other
one
was
I,
can't
remember
what
it
was.
I,
don't
think
it
was
a
connectome
paper,
but
this
one
there's
a
couple
of
papers
that
I'm
going
to
talk
about
now:
actually
neural
modeling
papers
and
C
elegans.
So
the
first
one
is
about
synaptic
organization,
and
this
is
actually
I.
A
Don't
know,
I
think
it's
explicitly
modeling
I
think
it's
more
biology,
but
it's
relevant
to
modeling.
This
is
about
synaptic
organizations.
This
is
a
bioarchive
preprint
that
came
out
earlier
this
year.
This
is
synaptic
organization,
the
C
elegans
neural
network,
which
is
of
course
the
C
elegans
connectome.
So
you
have
about.
You
have
302
cells
and
the
CL
against
connectome
and
they're
connected
in
different
ways,
they're
connected
either
through
Gap
Junctions,
which
is
the
most
common
way
to
represent
the
connectome
Gap
Junctions,
being
these
electrical
Junctions
between
cells
and
there's
a
fat.
A
What
they
call
fast
signal
there
and
that's
how
cells
commute
in
in
not
just
neurons,
but
all
cells
have
Gap
Junctions
and
in
in
neurons
at
least
they
have
these
fast
connections,
which
are
just
like
passing
information
through
the
Gap
Junctions,
but
they'll
you'll,
of
course,
also
have
synaptic
connections.
So
you
know
neurons,
don't
always
abut
one
another.
A
They
don't
always
sit
next
to
one
another,
sometimes
they're,
far
apart,
and
so
they
get
connected
by
extending
their
axons
out
and
then
extending
out
some
synapses
and
then
making
synaptic
connections
and
those
synaptic
connections
can
vary
in
larval
development
and
in
adulthood
depending
on
what's
going
on
in
the
environment,
so
sometimes
in
in
larval
development.
You
know,
if
there's
a
scarcity
of
resources,
the
worm
can,
you
know,
shut
down
certain
connections
or
it
can
emphasize
certain
connections
for
certain
behaviors.
A
So
neurons,
of
course,
are
characterized
by
tree-like
dendritic
structures,
which
are
the
the
cell
body,
and
then
these
dendrites
that
come
out
the
synapses
are
the
connections
between
those
those
dendritic
branches
and
another
neurons
branching.
You
know
so
it's
like
those
dendritic
branches,
Branch
out
from
the
cell
body
and
reach
out
to
other
cells
in
their
like
processes,
and
they
join
together,
and
this
is
this
enables
chemical
communication.
A
So
there's
there
are
a
lot
of
neurochemicals
that
get
passed
through
these
synapses,
so
they're,
very
diverse
in
terms
of
what
they're
enabling
in
terms
of
signaling,
so
so
neurons
are
characterized
by
these
dendritic
structures
that
support
local
computations,
but
integrating
multiple
inputs
from
Upstream
presynaptic
neurons.
So
you
have
a
presynaptic
neuron
and
a
postsynaptic
neuron.
A
It
is
less
clear
of
simple
neurons,
consisting
of
a
few
or
even
a
single
neurite,
which
is
the
neuride,
is
the
like.
The
cell
body
May
perform
local
computations
as
well.
So
these
are
simple
neurons,
as
opposed
to
you
know.
You
have
neurons
that
are
very
diverse
in
terms
of
their
signaling
in
terms
of
their.
A
Neurotransmitters
that
they
have
and
some
that
aren't
so
it
it's
really
variable
by
the
in
terms
of
the
actual
neuron
that
you're
talking
about.
So
they
want
to
know
about
local
computations.
To
address
this
question,
we
focused
on
the
compact
neural
network
of
C
elegans,
which
is
of
course
302
cells.
It's
always
302
cells,
there's
some
minor
differences
in
larval
development,
but
you
know
those
are.
A
We
know
what
those
cells
are
and
their
connections
to
address
this
question
we
focused
on
the
compact
neural
network
of
Clans
animals,
for
which
the
full
wiring
diagram
is
available,
including
the
coordinates
of
individual
synapses.
So
we
actually
know
that,
like
where
the
synapses
are
and
like.
If
we
take
an
image
of
a
c
elegans,
we
can
observe
the
synapses
as
well
as
the
cells.
So
a
lot
of
microscopy
you
can
identify
synapses
in
the
in
the
microscopy
data.
You
can
see
them.
They
look
like
little.
A
A
So
you
know
you
might
have
synapses
in
a
certain
area
where
you
want
to
perform
a
computation,
and
so
that's
that's
what
they're
finding
here
I'm
assuming
there's
some
images,
so
we
might
go
down
later
and
look
at
those
in
mutually
synapsing,
neurons
connections
of
opposite
polarity
clusters
separately.
So
these
connections
of
different
polarities
clustering,
different
clusters,
suggesting
that
positive
and
negative
feedback
Dynamics
may
be
implemented
in
discrete
compartmentalized
regions
along
the
rights.
A
A
Along
the
processes,
so
actually
each
cell
will
have
a
bunch
of
processes
coming
out
like
this,
and
this
will
be
just
you
know.
This
is
the
presynaptic
neuron
for
this
postsynaptic
neuron,
but
it
also
serves
as
the
presynaptic
neuron
to
B,
which
is
a
postsynaptic
relationship.
So
you
have
this
reciprocal
connectivity.
You
have
these
branching
structures
and
then
these
two
these
are
of
different
polarities,
but
these
synapses
are
all
grouped
here
in
a
cluster
and
so
spatially
they
they
exist
as
a
cluster,
and
so
that's
what
they're?
A
C
A
A
A
We
could
measure
if,
if
that's
but
they're,
all
different
connections
here
so
then
this
is
in
triple
neuron
circuits.
The
non-random
synaptic
organization,
May
facilitate
local
functional
rules
such
as
signal
integration
and
coordinated
activation
of
functionally
related,
Downstream
neurons.
So
there's
this
there
are.
These
different
types
of
this
is
still
non-randoms
inaptic
organization.
This
may
facilitate
local
functional
roles,
different
things
going
on,
and
these
clustered
synaptic
topologies
emerge
as
a
guiding
principle
in
the
network,
presumably
to
facilitate
distinct
Parable
functions
along
single
right,
effectively
increasing
the
computational
capacity
of
the
network.
A
A
You
can
actually
characterize
that
connection
with
the
number
between
zero
and
one.
So
a
lot
of
times,
they'll
have
a
strength
of
a
connection
or
sometimes
it'll,
just
be
either
a
connection
or
no
connection.
So,
that's
your
that's
how
you
usually
characterize
it
in
a
in
a
data
model,
but
they
want
to
actually
go
further
and
ask
you
know
if
we
look
at
like
some
of
these
synapses
that
cluster,
because
the
synapses
can
be
diverse,
they
can
have
different
functions.
They
might
pass
certain
neurotransmitters
over
others
depending
on
the
synapse.
A
It
may
allow
for
a
little
bit
more
diversity
in
terms
of
the
signaling,
so
so
okay
so
like,
for
example,
songbirds
Implement,
a
reliably
coincident
detector
based
on
nonlinear
summation
of
multiple
inputs
of
the
dendritic
tree.
So
you
have
all
these
different
inputs
coming
along
this
tree
and
the
question
the
neuron
has
to
sum
a
lot
of
those
things
or
they
have
to
sort
of
summarize
the
signal
so
you're.
Getting
all
these
different
signals.
A
You
know
a
in,
maybe
C
are
also
going
into
B
and
there
are
all
these
signals
and
they're
coming
into
the
cell
and
the
sauce
to
know
which
ones
are
the
ones
that
wants
to
act
upon
and
then
fire
and
then
move.
You
know,
move
the
information
down
to
the
next
neuron.
So
that's
why
we
care
about
like
these
different
synapse
clusters
and
this
computational
sort
of
nuance,
and
so
that's
so
they're.
You
know
you
see
things
in
songbirds.
A
A
People
have
done
things
in
like
they've
done
things
like
dendritic
computation,
so
dendritic
computation
is
where
you
have
a
cell
and
instead
of
having
these
inputs
with
weights,
you
have
all
these
different
signals
coming
in
and
then
the
signals
are
sort
of
captured
and
there's
some
rule
to
integrate
them.
So
it's
like
it's
almost
kind
of
like
this
model,
where
you
have
multiple
inputs
and
they're
coming
in
there's
some
rule
to
integrate
them,
so
they
use
them.
I
mean
they're,
more
experimental,
but
they've
used
them
for
different
types
of
signal
processing.
A
So
yeah
that's
and
then
so
in
the
absence
of
distinct
axons
that
output,
the
integrated
signal
from
the
dendrites
such
computations
may
be
performed
in
a
locally
compartmentalized
manner,
and
then
this
piece
of
local
activity
might
maybe
non-linearly
integrated
in
a
compartmentalized
Manner
and
then
transmitted
to
the
postsynaptic
neurites
within
the
compartment
and
without
evoking
current
changes
across
the
entire
neuron.
So
this
is
where
you
have
them
all
coming
in
one
place,
the
signal
gets
integrated
and
then
it
doesn't.
You
know
there
isn't
any
sort
of
variation
across
the
neuron.
A
If
you
had
things
coming
in
from
different
from
different
dendrites
coming
into
the
neuron
into
the
cell
body,
you
know
you
might
have
like
local
differences
across
the
cell
in
terms
of
all
kinds
of
information,
because
if
you
integrate
it
locally,
then
you
have
these
when
they
get
when
it
gets
converted
into
some
sort
of
electrical
activity
there
can
be
or
changes
in
current.
So
you
have
these
changes
in
you.
A
Have
these
neurotransmitters
they're,
invoking
these
changes
in
current
and
then
they're
being
translated
into
some
sort
of
electrical
activity
in
the
neuron.
So
it's
kind
of
an
interesting
relationship
and
it
kind
of
speaks
to
Susan's
question
about
what's
going
on
with
local
charge
densities
and
the-
and
there
may
be
that
you
know
you
have
these
currents,
you
have
this
balance
of
different.
A
You
know
you
have
this
sort
of
regulation
of
the
current
within
neurons
and
then
that's
relating
to
some
sort
of
functional
electrical
activity
later,
so
this
is
a
sort
of
a
better
drawing
than
what
I
did
there.
On
my
board,
you
have
the
inputs
here
on
a
dendritic
tree,
and
this
is
like
you
see
this
dendritic
computation
as
well
as
in
a
biological
system.
You
have
these
inputs
to
this
tree,
which
is
basically
the
branches
that
bring
information
or
bring
this
information
into
the
neuron.
A
Cell
and
then
there's
some,
you
know
there's
some
activity
in
the
cell
that
generates
an
output.
It
generates
it.
Maybe
an
action
potential
theologism
doesn't
really
have
and
we'll
talk
about
this
in
the
next
paper.
It
doesn't
have
true
Action
potentials,
but
put
that
aside
for
now
and
then
there's
an
output
to
the
next
cell,
which
gives
it
information.
A
It
gives
a
neurotransmitters
to
transmit
and
things
like
that
and
then
cells
can
of
course
become
entrained
with
one
another
throughout
these
Networks,
and
this
is
of
course
the
compartmentalized
computations
they're
talking
about
where
you
have
these
clusters
of
inputs
and
there's
an
output
here
where
it
may
just
go
from
this
input
to
the
next
cell
or
it
may
go
down
to
the
cell
body.
So
there
that's
that's
the
benefit
of
these
input,
clustered
inputs.
A
So
they
do
this
they've
actually
compiled
the
database
of
synaptic
coordinates,
which
is
they're
using
connectome
data
provided
by
reference
15,
which
is
they
don't
have
the
paper
pile
or
fighter,
but
they
contain
the
skeleton
maps
of
most
neurons,
as
well
as
the
position
of
chemical
and
electrical
synapses
across
along
the
neurons.
So
I
think
these
are
like
microscopy
images
but
skeleton
maps
of
those
microscopy
images.
So
they
have
this
information
that
is
available
and
there's
like
they
have
a
coordinate
system
for
it.
A
A
A
So
having
these
double
I
mean
we
know
the
connectome
of
the
hermaphrodite
and
there's
minimal
difference
between
that
and
the
male,
but
having
those
connectomes.
You
know
it
just
gives
us
a
better,
better
resolution
on
C
elegans.
So
that's
where
they
got
their
data
from
here,
but
there
are
other
like.
If
you
don't.
You
know
there
are
other
data
sets.
If
you
want
to
look
at
the
connecting
them.
I
know.
I
talked
to
Alana
about
this
they're.
A
host
of
other
data
sets
that
have
and
I
put
them
in
this.
A
In
the
slack
Channel,
if
you're
interested
in
that
further-
or
we
can
talk
about
it
further
if
you're
interested,
so
this
is
a
figure
here
where
they,
they
kind
of
talk
about
these
three
neuron
circuits
connected
by
clustered
synapses
that
may
support
local
functional
roles.
So
in
a
you,
have
this
three
possible
layouts
for
a
pair
of
neurons.
A
Think
in
my
word,
I
came
up
with
a
little
bit
different
Arrangement,
but
you
know
they
just
they
consider
these
three
and
then
they
have
this
where
they
have
two
neurons
here,
or
at
least
I,
think
the
branching
from
these
two
neurons
and
they
match
up
with
these
synapses.
So
these
are
like
this
is
this
diagram
is
a
little
confusing,
but
so
this
is
a
common
neurite
C.
A
So
this
is
Neuron
C
in
this
network,
and
these
are
the
synapses
that
exist
along
this
projection
and
they're
clustered
like
so,
and
you
can
see
that
like
Neuron,
a
and
neuron
B
also
have
their
processes
and
they
need
to
connect
somehow
to
see
so
they're.
Actually,
these
connections,
you
can
see
ac1
bc1
ac2
bc2,
so
you
see
that
there
are
these
common
connections
so
like
a
is
connecting
to
C
here,
B
is
connecting
to
C
here.
A
This
is
a
and
C
another
connection,
but
you
can
see
that
they're
making
connections
in
these
clusters
on
C,
so
A
and
B
are
connecting
projecting
to
the
neurot
of
C
that
those
connections
are
are
based
on
cluster.
You
know
they're
projecting
the
Clusters,
and
so
then
you
compare
them
in
different
ways.
So
it's
a
combinatoric
system
where
you
have
different
clusters
here
of
synapses
that
are
paired
between
two
cells
and
their
projections,
and
then
you
can
actually
calculate
a
mean
distance
between
these
different
pairings.
A
So
you
can
calculate
a
shuffled
mean
distance,
so
you
can
actually
look
at
it
like
a
randomized
version
of
this
to
assume
that
maybe
there
isn't
like
clustering
or
functional
clustering
involved
in
this-
and
you
can
look
at
those
two
things
and
in
comparison
this
is
number
is
C
is
actually
probably
a
map
of
Okay.
So
that's
all
a
so
long,
Legend
an
example
for
tightly
clustered
set
of
synapses.
So
this
is
C.
You
have
three
neurons
are
riml
or
mdl
Rd
rmdr,
and
these
are
all
neurons
and
C
elegans.
A
This
is
the
nomenclature
they
use,
and
then
this
is
just
these
colored
lines
or
the
connections
between
the
two.
So
the
red
is
our
raml
to
rmdl.
So
this
is.
These
are
two
neurons
on
the
left
side
of
this,
of
the
worm
where
you
have
rim
and
rmd
and
they're,
both
left
on
the
left
side,
and
they
have
connections
raml
and
rmdr.
So
this
is
a
bilateral
Connection
Blue,
and
so
you
can
see
them
here
and
then
other
synapses
shown
in
Gray.
So
you
can
see
these
blue
dots
here
that
are
clustered.
A
You
see,
the
red
dots
are
clustered
and
you
see
the
gray
dots
which
are
up
and
down
it's
very
hard
to
see
here,
but
they're
up
and
down
these
projections
and
they're
not
clustered.
So
you
can
see
how
those
relationships
sit
with
real
respect
to
real
neurons,
so
I
think,
that's
probably
all
I'll
say
about
this
paper.
It's
a
nice
paper.
If
you
want
to
read
it
further,
they've
worked
this
all
out
and
they
have
a
nice
model.
A
This
is
actually
a
nice
image
might
as
we'll
get
into
this
of
some
of
the
Imaging
data
that
they
have.
Here
we
go
so
this
is
compiling
their
database.
This
is,
this
is
a
ring
here
of
nerves,
typical
nerve
ring,
and
this
is
some
of
the
how
what
these
look
like.
So
these
are
synapses.
We
have
you
know
different
types
of
neurons
in
the
nervous
system
of
C
elegans
we
have
motor
neurons,
Sensory,
neurons
and
interneurons,
which
sit
between
the
motor
and
sensory,
neurons
and
they're
all
color
coded
here.
A
A
This
is
the
nerve
ring
here
which
is
in
the
you'll,
see
this
in
the
C
elegans
along
the
body
axis,
and
this
is
again
is
the
body.
This
is
in
context
where
it's,
this
nerve
ring.
You
have
a
lot
of
connections
here
along
these
different
projections
from
different
types
of
neurons,
and
so
that's
where
this
connectivity
resides,
and
so
this
is
also
how
you
can
get
your
data.
You
can
get
it
from
microscopy
data
where
the
neuron,
where
the
synapses
are
marked
in
the
images.
A
So
this
is
a
where
they
have
the
marked
you
can
identify
them
in
the
in
the
microscopy.
This
is
one
micron,
so
this
is
a
pretty
high
resolution
image
and
they
can
actually.
This
is
I,
think
a
close-up
of
some
area
where
they've
been
able
to
mark
this.
This
is
a
synapse
of
our
APR
and
uradr.
So
this
is
where
you
can
tell
where
those
two
connect
and
then
they've
done
a
distance
between
neuroid
centroids,
so
they
actually
have
a
map
of
the
distance
between
the
centroids
and
they
have
so.
A
So
that
means
how
far
are
the
synapses
out
from
the
centroid
of
the
cell
on
these
projections
and
they
find
that
in
general
they're
not
very
far
out
in
terms
of
this,
this
Micron
measurement
so
they're
using
the
distance
in
microns
and
it's
actually
not
even
a
micron,
and
so
there's
this
distance.
Actually
it's
between
the
projection,
it's
not
between
the
cells
but
the
centroids
of
each
one.
So
you
can
see
here
that
there's
a
average
distance
so.
A
To
build
a
atlas
from
that
those
data,
so
that's
that's
how
they
do
that
second
paper
I
wanted
to
talk
about.
Is
this
paper
on
modeling
of
three
types
of
non-spiking,
neurons
and
C
elegans?
So
I
talked
about
how
this
is
C.
Elegans
doesn't
really
have
traditional
action.
Potentials
like
you
might
see
in
a
human
neuron
or
some
a
million
neuron,
or
something
like
that
they
have
a
different
type
of
not.
They
have
non
different
types
of
non-spiking,
neurons
and
I'm.
A
Not
sure
the
reason
for
this,
but
this
this
paper
goes
through
some
of
the
biology
here.
This
is
the
international
Journal
neural
systems.
This
is
back
in
2005,
but
it's
a
nice
review
of
these
different
types
of
non-spiking
neurons.
So
the
abstract
is
reads.
The
nematode
C
elegans
is
a
well-known
model
organism
Neuroscience.
The
relative
Simplicity
of
its
nervous
system
share
some
essential
features
of
the
more
sophisticated
nervous
systems.
A
We
need
to
fully
characterize
the
nervous
system
following
a
recently
conducted
electrophysiological
survey
on
different
ceiling
in
the
neurons.
This
paper
aims
at
modeling
three
non-spiking
neurons,
and
this
is
these-
are
Ram
a
a
I
y
and
afd,
and
these
are
just
like
the
nomenclature
for
different
types
of
neurons,
so
yeah.
The
convention
is
the
name
used
three
uppercase
letters
and
then
the
last
letter,
as
usually
like
the
left
or
right
side.
So
that's
how
they,
usually.
A
These
are
adult
neurons
that
have
different
terminally
differentiated
and
that's
that's
how
they
do
this
this,
and
these
these
conventions
are
different
than
the
developmental
cells
which
you'll
see
in
a
lineage
tree
where
they
have
like
their
sub
lineage.
And
then
the
letters
like
you
know
left
right
or
dorsal
ventral
or
something
like
that.
They'll
have
like
they
have
a
totally
different
convention
for
the
adult
cells.
A
To
date
they
represent
the
three
possible
forms
of
non-spiking
neuronal
responses.
So
these
are
just
three
examples
of
these
three
different
types
of
neurons.
To
achieve
this
objective,
we
propose
a
conductance-based
Neuron
model
adapted
to
the
electrophysiological
features
of
each
neuron.
These
features
are
based
on
current
biological
research
and
a
series
of
insilical
experiments,
which
is
differential
Evolution
to
fit
the
model
to
experimental
data.
We
formulate
a
series
of
hypotheses
regarding
currents
involved
in
the
neurodynamics.
A
These
models
reproduce
experimental
data
with
a
high
degree
of
accuracy
while
being
biologically
consistent,
and
so
they
kind
of
go
through
these
examples,
and
so
this
is.
These
are
the
three
here
three
different
types
of
neuron:
actually
they
have
four
here
they
have
Ram
transient
outward
rectifier
aiy,
outward
rectifier
afd
by
stable
and
then
AWA
bistable
spiking,
so
that
those
are
the
four
that
they
have
in
this.
A
A
You
know
this
is
electrical
activity
in
millivolts
and
you
can
see
that
how
that
happens-
and
this
is
so-
these
are
just
the
characterize.
The
four
forms
of
possible
neuronal
response
in
the
nematode
and
so
figuration
is
the
evolution
of
the
membrane
potential
from
a
series
of
current
injections,
which
is
just
a
a
technique.
They
use
to
sort
of
find
the
electrical
activity
or
to
elicit
it
and
record
figure.
B
represents
the
evolution
of
total
ion
currents
of
different
neurons
when
their
membrane
potentials
are
clamped
at
a
fixed
value.
A
A
So
this
is
a
lot
of
neurophysiology
that
I'm
not
really
had
a
huge
amount
of
background
in
and
I've
I've
become
a
little
bit
familiar
with
it,
but
not
that
familiar
so
I
don't
want
to
pretend
to
be
an
expert
at
this,
but
they
they
definitely
have
like
some
interesting
aside
from
sort
of
the
interesting
nuances
of
C,
elegans,
aerophysiology,
they're
kind
of
you
know
just
kind
of
interesting
systems.
We
know
those
you
know,
cell
by
cell.
We
know
what
they
kind
of
do.
They
know
what
their
function
is.
A
So
it's
an
interesting
system
to
look
at
this
and
see
what
kinds
of
electrophysiological
activity
occur.
So
so
yeah
they
talk
about
different
ion
channels
that
exist
in
some
cells,
but
not
others.
So
you
know
you
have,
for
example,
calcium
ion
Channel
some
papers
show
the
existence
of
a
calcium
ion,
Channel,
Dynamics
and
Ai,
and
efd
neurons,
as
well
as
RAM
neurons.
Additionally,
calcium
currents
have
been
reported
in
similar
neurons,
such
as
AWA
neurons,
AWC,
neurons,
aser,
aib,
Ava
and
Ria.
A
So
there's
like
this
calcium
ion
Channel
that's
been
reported
in
some
some
cells
and
not
others.
The
presence
of
an
inwardly
rectifying
potassium
current.
So
when
I
talked
about
those
different
classes,
those
different
classes
imply
that
they
have
different
sets
of
ion
channels
that
are
active
in
these
cells,
and
so,
in
this
case,
there's
an
inwardly
rectifying
potassium
current,
which
is
a
potassium
ion.
Channel
has
been
experimentally,
confirmed
in
AWA,
neuron
and
HSN
neuron,
then
there.
A
So
there
are
these
differences
in
C,
elegans
neurons,
both
amongst
just
C,
elegans,
neurons
and
within
C,
elegant
neurons,
so
yeah.
So
that's
and
then
they
did
this
differential
Evolution
where
they
look
at.
They
use
a
population-based
meta
heuristic.
So
this
is
an
evolutionary
algorithm
approach
which
is
nice.
A
They
kind
of
work
on
that
as
a
as
a
way
to
do
optimization.
So
this
is
what
they
call
Meta
heuristic,
which,
if
you're
computationally
Savvy
you
know.
That
is
something
that
you
might
use
as
a
way
to
kind
of
characterize:
optim,
optimal
Behavior
across
the
large
set
of
things,
and
you
can
use
in
evolutionary
algorithms.
These
populations
of
of
these
different
things
as
a
way
to
or
we
evaluate,
the
population
based
on
a
fitness
function.
So
so
they
try
to.
A
They
try
to
create
Solutions
using
differential,
Evolution
and
find
Optimal
values
for
this.
So
they
actually
have
their
code
on
GitHub
parameter.
Estimation
for
using
differential
evolution
and
they're
able
to
actually
come
up
with
some
parameters
from
the
literature
and
they
have
minimum
maximum
values
for
their
simulation.
And
then
they
do
the
simulation.
A
They're
able
to
actually
model
some
of
these
neurons.
There
are
some
values
for
some
of
these
neurons,
given
things
from
the
literature
and
things
that
they've
been
able
to
observe
for
these
for
these
cells.
So
this
is
for
REM
a
I
y
and
afd
okay
presented
in
that
table,
and
they
do
you
know
some
other
statistical
tests.
They
look
at
the
evolution
of
our
membrane
potential,
so
I
think
the
take-home
message
of
this
paper
is
that
there's
a
diversity
of
cells
within
C
elegans
connect
Dome.
A
A
C
A
C
B
C
B
B
C
Yeah
because
they
have
volume-
yes,
okay,
that's
I'm,
just
my
I'm,
just
considering
single
cells
and
then
making
a
grouping
of
them.
It's
kind
of
starting
with
the
single
one
and
then
I've
already
bought
this
drawn
out,
but
it's
unstable.
It
throws
I,
it
says:
eigenvectors
are
what
was
it?
It's
eigenvectors
are
bad
I
got
a
negative
or
something
please
show
me
the
show
me
the
math.
Then
you
stupid
program
I
figured
out
how
to
find
the
matrices
that
go
with
the
computations
and
I
asked
the
people
from
console
for
that.