►
From YouTube: DevoWorm #41: Making a Statistical Mask, Origins of Early Life, Minimal Cells, Autocatalytic Sets
Description
Creating a statistical mask from distributions. Modeling the origins of life with Braitenberg Vehicles. Mathematical modeling of autocatalytic sets for innovation and origins. Minimal cells and simulating whole bacterial cells, putting cellular components together into a whole. Attendees: Richard Gordon, Morgan Hough, Susan Crawford-Young, Jesse Parent, Bradly Alicea, and Morgan Hough
B
B
A
B
B
A
A
A
B
A
A
I,
it's
I've
been
doing
some
stuff
with
some
of
the
things
that
we've
been
working
on.
I'll
share
my
screen
on
this,
so
the
first
thing
I
have
is
this
demo
of
embedding
elements
and
noise
I'm
not
sure
this
is
what
we're
thinking
about,
but
I
was
working
with
scilab,
which
is
a
program,
that's
kind
of
like
Matlab
you,
you
know,
but
it's
open
source.
So
you
can
work
on
matrices
and
transforming
matrices
and
I've
been
doing
this
noise.
A
A
B
A
You
can
put
in
any
set
of
values
for
The
tensegrity
Matrix,
but
it's
usually
you
know,
I
mean
you
can
make
comparisons
between
matrices.
So
if
you
say
this,
Matrix
represents
the
structure,
and
this
is
these.
Are
the
forces
and
you
get
like
something
at
the
end?
That's
a
that's
like
a
rank
of
The
Matrix,
and
so
it
it
gives
you
a
measure.
A
A
Certain
aspects
of
the
physics
so
like,
if
there's
something,
there's
a
there's,
a
dynamic
tension,
it's
like
one,
if
it's
neutral,
it's
zero
and
if
it's
not
it's
negative
100.
So
it's
like
at
each
node,
there's
a
certain
property
to
that
node.
If
it's
in
Dynamic
tension
or
not
I
can't
remember
other
Criterion,
but
that
I
mean
it's
very
pretty
easy
to
follow
through
you
just
put
some
put
in
a
matrix
and
you
can
get
a
a
measure
at
the
end
and.
B
B
A
B
A
So
this,
okay,
let
me
go
through
this
yeah,
so
this
is
a
poisson
distribution.
This
is
actually
just
you
know:
defining
a
poisson
distribution
in
terms
of
its
gamma
value,
which
is
kind
of
like
it's
descriptive,
parameter
and
higher
the
descriptor
parameter.
The
more
points
you
get
going
like
kind
of
up
so,
like
you
know
the
lower
the
gamma
it
it
kind
of
resembles
gaussian
distribution,
the
higher
the
gamma
you
get
more.
A
A
So
this
is
the
poisson
signal
that's
out
in
the
world,
and
then
this
is
where
there's
some
noise
uniform
noise,
that's
generated
these
Green
Dots,
and
so
these
are
generated
and
I
can
plot
them
over
the
poisson
to
hide
those
points.
So
the
blue
points
are
my
poisson
points
that
I
generated
in
this
step,
and
then
here
I've
generated
Some,
Noise
overlaying
this
these
poisson
points,
and
so
it's
hiding
some
of
the
outlier
points
here.
A
A
A
But
yeah
yeah
there
are
a
lot
of
them
there
so
and
then
so,
they're
just
circles,
and
then
these
are
also
circles
he's
green,
but
because
there's
so
many
of
them.
You
know
you
get
this
field
of
it's
like
a
green
field
over
these
blue
circle
lines
and
then
sort
of
embeds
it
and
you
you
know,
I
was
like
playing
around
with
the
aspect
ratio
of
the
image,
and
so
you
can
get
like
more
crowding.
A
C
A
So
it's
it's
just
that
it
goes
from
left
to
right.
So
it's
generating
things
in
time
and
the
idea
is
that
it's
like
they
call
them
arrivals,
sometimes
in
in
poisson
distributions.
But
if
you're
measuring,
like
a
calculation
of
how
many
things
You
observe
at
a
certain
time.
So
it's
like,
if
you
think
about
like
a
highway
where
you're
standing
in
an
overpass
and
you're
okay,
how
many
cars
pass
in
like
a
minute
and.
B
C
A
B
A
A
Distribution
of
points
and
then
that
that's
overlaying,
the
poisson
distribution
so
you're
hiding
some
points,
they're
less
frequent
in
the
poisson
distribution,
but
the
more
frequent
points
that
occur
are
you
can
see
through,
and
so
the
point
of
this
then
is
to
have
like
this
mask
over
the
the
distribution
with
the
uniform
mask
that
you
know
kind
of
hides
some
of
those
points.
And
then
that's
that's
what
I
guess
we
wanted
for
detecting
points
out
of
the
back
road.
C
A
A
A
C
C
C
A
A
A
They
did
I
was
looking
into
purlin
noise,
they
don't
have
it
in
the
grand
function
and
it's
a
little
bit
more
complex,
but
I
I
have
I
think
it
might
work,
I'm,
I'm
working
on
it
inside
lab.
It
would
be
nice
if
it
did,
because
you
could
actually
run
it.
It's
a
procedural
thing,
so
it
would
require
like
building.
C
C
A
A
And
this
is
where
it
now,
this
distribution
that
you
see
here,
which
is
a
poisson
distribution,
that's
been
sort
of
permutated
and
transformed,
so
that
most
of
the
signal
is
down
at
the
bottom.
So
it's
from
that
distribution
between
permutated
and
so
now,
I
have
a
band
at
the
bottom.
Where
you
can
see
it's
a
little
bit
more
visible
and
you
can
do
things
like
that
or
you
can
do
another
one
where
there's
a
band
at
the
top.
It's.
C
A
A
We
I
we've
gone
over
it
a
little
bit
and
I'm
going
to
go
over
the
paper
on
you
know:
I've
been
working
on
the
bradenberg
vehicle
section
kind
of
making.
I
don't
have
that
version
on
on
this
open
here,
but
I'm
working
on
some
of
the
details
in
terms
of
how
this
might
work,
how
we
might
be
able
to
like
the
parts
list
and
and
how
to
make
that
sort
of
more
Salient,
because
we
had
I
know
I
sent.
B
A
C
Okay,
good
yeah,
okay,
yeah
that'll
be
interesting
to
see
how
many,
how
many
random
structures
actually
function:
yeah,
yeah,
okay,
okay,
I'm,
not
sure
what
the
Criterion
is,
because
you
know
with
living
organisms.
Let's
say
microorganisms:
you
can
get
weird
mutants
that
go
around
in
circles
or
make
patterns
and
do
all
sorts
of
things
other
than
swim.
Normally
right
and
I
guess
you'd
still
call
them
alive,
yeah,
it's
just
that
they
they
have
a
different
function
in
life.
Yeah.
C
A
C
Know
that
you
know
like,
like
we
heard
a
crazy
guy
who's,
does
the
he
throws
himself
to
the
magnetic
reversal
group,
but
with
Magnetic
Universal
what
you
call
it
like
that
start,
a
podcast,
because
you
can
see
them
yeah,
but
he
was
reporting
on
a
bunch
of
sheep
that
go
around
in
circles
all
the
time.
Oh
yeah,
no
he's
not
he's
attributing
this
to
changes
in
the
magnetic
field,
whereas
it's
a
well-known
disease
in
sheep.
C
C
C
A
As
a
follow-up
to
this
stuff
on
early
life
that
we're
talking
about
I'd
like
to
go
over
a
paper
on
auto
catalytic
sets,
so
this
is
a
paper
from
Stuart
Kaufman,
William,
hordyke
and
Mike
Steele.
These
are
some
pretty
big
names
in
like
complexity,
Theory
and
modeling,
and
this
paper
is
called
Auto
catalytic
sets
arising
in
a
combinatorial
model
of
chemical
Evolution.
So
this
ties
to
the
origin
of
life.
This
is
talking
about
chemical
evolution
in
the
origin
of
life.
A
A
You
have
chemicals
in
the
environment
of
an
early
Planet,
this
case
early
Earth.
How
did
they
come
together
to
build
coherent
and
self-reproducing
collectives?
And
so
that's
the
question
they're
answering
here.
One
possible
answer
was
proposed
in
the
form
of
emergence
of
an
auto
catalytic,
set,
a
collection
of
molecules
that
mutually
catalyze
each
other's
formation.
That
is
self-sustaining
and
gives
some
basic
food
source.
A
So
you
need
a
couple
of
things
you
need.
One
is
that
Mutual
catalyzation,
because
you
can't
have
a
master
catalyzer
when
a
system
is
self-organizing
and
something
that's
self-sustaining
through
some
sort
of
food
source.
So
there's
got
to
be
some
sort
of
metabolic
input
building
on
previous
work.
Here
we
investigate
in
more
detail
when
and
how
Auto
catalytic
sets
can
arise
in
this
simple
model
of
chemical
evolution,
so
they're
doing
this
modeling
exercise
to
show
how
these
might
arise.
B
A
Catalysis
assignments:
this
is
just
kind
of
a
Innovation
model
that
they
use.
We
derive
theoretical
results.
We
compare
them
with
computer
simulations,
so
this
is
a
lot
of
these
early
life.
Experiments
using
chemical
Evolution
were
done
by
little
parent
and
holding.
Of
course.
These
two
people
are
very
important
in
the
early
in
in
the
origin
of
life
in
an
evolutionary
biology,
so
that
they've,
you
know
these
are
Big
names
in
those
areas.
There's.
A
A
Acids
under
possible
primitive
Earth
conditions
where
they
took
a
broth
of
different
molecules,
they
put
an
electric
charge
across
them
and
they
produced
a
lot
of
different
metabolites
and
they
redid
the
experiments
more
recently
back
in
2015,
actually
a
little
bit
earlier
than
that,
and
they
found
that
they
got
a
much
richer
array
of
of
biomolecules
to
result
from
it.
So
the.
A
C
A
It's
just
something
to
think
about
we're
thinking
about
early
life.
This
paper
actually
goes
through
a
number
of
different
computational
models
or
theoretical
models
for
understanding
this
citizen
experimental.
The
first
method
is
the
tap
model.
The
tap
model
was
based
on
the
idea
of
combinatorial
innovation,
which
is
where
you
have
a
number
of
different
combinations
that
can
be
formed,
and
this
leads
to
innovation
in
certain
ways.
So
if
you
recombine
things
over
and
over,
you
get
innovations
that
emerge,
and
so
they
use
this
equation
here.
A
Where
m
m
sub
T
is
the
number
of
different
things
at
time.
T
and
Alpha.
Sub
I
is
the
decreasing
sequence
of
probabilities.
So
this
is
again
with
this
stuff,
where
we
talk
about
regenerating
a
bunch
of
model,
a
bunch
of
configurations
of
different
probabilities,
this
this
combinatorial
model,
the
tap
model
actually
probably
would
do
well
to
describe
what
we're
trying
to
do
with
the
Brandenburg
vehicles
and
so
interpret
a
new
context
to
chemical
evolution.
New
molecular
species
are
produced
their
chemical
reactions
of
arbitrary
combinations
of.
B
A
Is
exactly
the
kind
of
thing
that
happened
in
the
military
experiment,
except
it
happened
in
great
parallelism
and
you
know
you
get.
You
have
a
basically
a
huge
broth
of
possibilities
there
now
this
above
equation,
this
m
sub,
t
plus
one
and
solving
for
that-
represents
a
deterministic
version
that
does
not
guarantee
M
sub
T
to
be
an
integer
value
and.
A
The
general
idea
behind
the
model,
so
these
are
stochastic
implementation
for
their
simulations
here
for
each
newly
created
molecular
species
X
in
each
of
the
existing
molecule
types
y
that
were
not
present
at
Time
Zero
at
scheme
catalyze
the
formation
of
Y
with
a
fixed
probability
P.
Similarly,
for
each
chemical
reaction,
r
that
produces
a
new
species
and
each
of
the
existing
molecule
times
y
that
were
not
present
at
Time
Zero
y
can
catalyze
R,
also
with
probability
P.
These.
B
A
Through-
and
they
do
this-
that
there's
an
algorithm
tapped
with
catalysis,
they
give
the
algorithm.
This
is
pseudocode.
It
basically
describes
this
process,
then,
over
time,
the
M
sub
T
existing
molecular
species
in
the
particular
chemical
reactions
and
produce
them
thus
form
a
growing
chemical
reaction,
Network,
which
is
shown
here,
there's
a
food
set
containing
of
consisting
of
the
M
suppose
initial
species,
as.
A
Assignments
and
the
dashed
arrows
and
this
feed
forward
network
with
the
solid
arrows.
So
this
is
over
time.
This
is
over
M
Sub
0
m
sub,
1
M
sub
2
m
sub.
Three,
you
end
up
at
10.,
so
that's
what
their
Network
looks
like.
They
also
have
RAF,
which
is
now
the
probability
that
such
a
growing
chemical
reaction
Network
and
some
at
some
point,
a
subset
of
molecule
types
exist
in
which
the
molecule
mutually
catalyze,
each
other's
production,
which
is
sustainable
on
the
given
food
set.
So.
A
Is
this
idea
of
reflexively
Auto,
catalytic
and
food
generated
set
graph?
More
formally
graph
is
a
set
of
our
chemical
reactions
and
the
molecule
types
involved
in
them
such
that
number.
One.
Each
reaction
in
R
is
catalyzed
by
at
least
one
of
the
molecules
involved
in
our
and
two
each
molecule
type
involved
in
R
can
be
created
from
the
food
set
through
a
sequence
of
reactions
from
R
itself.
So
then
they
use
a
some
sort
of
algorithm
to
find
such
refsets
in
an
arbitrary
chemical
reaction.
Network
and
we're
determine
the
no
subset
is
present.
A
A
Generated
molecules,
and
is
the
union
of
all
possible
graphs
within
a
given
Network,
so
then
tap
and
wrap
can
be
used
in
tandem.
You
can
look
at
graph
sets
in
the
tap
model,
and
this
is
this
was
done
in
the
context
of
technological
evolution
in
21.,
which
is
this
paper
here.
Dynamics
of
a
birth
death
process
based
on
combinatorial
Innovation,
and
this
is
where
you're
looking
at
the.
A
Death
of
combinatorial
Innovations,
so
you're,
looking
at
like
things
that
innovations
that
die
and
are
born
and
the
process
there
so
you're
you're
getting
you
can
apply
this
model
to
other
things
other
than
the
origin
of
life.
But
in
this
case
they
apply
to
the
origin
of
life
and
they
do
a
lot
of
modeling
here
for
different
parameter
values
and
show
that
this
model
is
fairly
effective
at
doing
what
they
want
to
do.
A
A
A
C
A
C
C
A
B
B
C
C
B
C
A
C
C
A
B
A
A
And
they
they
did
a
they
did
this
at
a
conference.
This
summer
we
talked
about
it
and
they're
doing
this
stuff,
with
This
Is
Fundamental
behaviors
emerging
simulations
of
a
living
minimal
cell.
So
this
is
where
you're
building
a
minimal
cell
you're,
taking
out
like
everything
that
doesn't
isn't
required
for
the
cell
dysfunction
and
then
trying
to
build
a
cell
that
has
like
the
bare
components.
A
So
the
oftentimes
like
in
bacterial
cells,
will
take
out
all
the
genes
that
aren't
essential
and
they'll
have
this
list
of
essential
genes,
and
you
might
want
to
do
that
say
to
like,
if
you
want
to
build
a
bacterial
cell,
that
can
just
function,
metabolically
and
then
booted
up
with
genes
to
do
something
like
you
know,
some
some
task
that
they
want
the
cell
to
do
so,
that's
why
you
might
want
a
minimal
cell.
This
is
the
basic
metabolic
requirements
of
a
cell
without
any
other
function
really.
A
C
C
A
C
C
A
A
Yeah
yeah,
then
they
have
these
receptors
here
and
ribosomes.
So
they're
like
they're,
looking
at
kinetic
data
from
experiments,
they're
going
to
get
cellular
processes,
then
when
they
get
that,
like
they
have
the
biological
side,
then
they
build
these
whole
cell
simulations,
which
are
these
dynamical
systems
that
predict
the
cell,
what
the
cell
does
and
then
they
they
compare
it
with
the
biology.
So
you
get
a
good
comparison
of
a
simulation
with
the
biology,
so
their
simulations
are
highly
detailed
and
highly.
You
know
like
they.
A
From
the
bio
biology
of
the
simulation
yeah,
so
they
do
this
GP,
they
run
it
on
a
GPU,
the
simulation
and
it's
you
know,
they're
able
to
extract
different
parameters.
So
these
time
dependent
cell
States,
looking
at
mRNA
cell
doubling
time
and
ATP
cost
so
yeah
so
yeah.
This
is
a
3D
spatial
resolution
of
a
fully
dynamical
wholesale
kinetic
model.
A
Then
you
have
these
single
reactions
that
are
pretty
detailed.
You
have
time
these
time
dependent
cell
State
measurements,
which
are
of
course
this
function.
Here
you
have
I
talked
about
mRNA
half-lives
and
looking
at
like
a
process
where
you
know,
if
you
have
a
transcriptional
process,
you
can
look
at
the
MRNA
half-life
of
that
process.
So
if
something
is
transcribed,
you
know
how
long
does
the
MRNA
live
in
the
cell
and
I've
done?
A
Experiments
like
this,
where
you
can
shut
off
the
mechanism
and
see
what
the
MRNA
Half-Life
is,
and
it
tells
you
something
about
the
underlying
mechanisms
of
metabolism
and
then
connections
among
metabolism.
Genetic
information
and
cell
growth
are
revealed,
so
they
actually
can
do
a
lot
of
these,
not
just
like
looking
at
a
minimal
genome.
It's
looking
at
the
whole
cell
and
how
it's
behaving
so
yeah,
let's
see
so
they
have
this
whole
cell,
fully
dynamical
kinetic
model
which
they
call
wcm
wholesale
model
of
jcbi
sim3a,
which
is
this
minimal
cell.
A
So
that's
not
that's
not
trivial.
It
requires
a
lot
of
computational
power,
but
that's
part
of
the
simulation
as
well.
This
reveals
how
the
cell
balances
demands.
The
cell
balances
demands
of
its
metabolism,
genetic
information
processes
and
growth.
So
there's
this
they
want
to
see
how
these
things
are.
A
You
know
basically
what
the
cell
is
investing
in
each
of
these
things
and
offers
insights
into
the
principles
of
life
for
this
minimal
cell.
The
energy
economy
of
each
process,
including
active
transport
of
amino
acids,
nucleosides
and
ions,
is
analyzed.
Wcm
reveals
how
emerging
imbalances
lead
the
slowdowns
and
the
rate
of
transcription
and
translation.
A
A
When
you
have
transcription
you're
measuring
it
from
the
production
of
mrnas,
you
can
watch
the
sort
of
the
concentration
of
those
and
you
know
for
in
terms
of
slowdowns.
The
mrnas
have
a
certain
Half-Life,
so
they
survive
in
the
cell
and
if
the
number
of
our
marinas
Falls
the
concentration
of
them
Falls,
and
you
can
assume
that
that
function
has
been
shut
down
or
diminished
in
some
way.
And
if
it's
you
know,
if
there's,
if
there's
a
higher
concentration
of
mrnas,
you
assume
that
that
process
is
enhanced.
A
A
A
So
these
are
all
things
that
the
you're
just
looking
in
the
experimental
system
and
you're
using
micro
mycoplasmas
in
this
case
in
this
paper-
and
so
this
is
a
very
simple
cell-
and
you
know
it's
very
easy
to
manipulate
and
to
observe.
So
that's
that's
why
they
use
micro,
mycoplasma,
and
so
we
have
been.
We
have
been
able
to
produce
living
cells
with
fewer
genes
than
any
known,
naturally
occurring,
so
so
this
is
some
of
their
work.
A
A
B
A
C
B
A
B
B
A
A
C
B
A
B
A
A
different
you
know
different
setups,
whereas
so
these
are
chemical,
Master
equation,
they
work
from
that
and
so
a
lot
of
times.
You
will
assume
that
the
cell
is
well
stirred
and
you
know
it
includes
diffusion
and
rates
of
transcription,
and
so
it
has
all
these
different
rates
involved
in
it,
but
it
doesn't
require
any
sort
of
spatial
segregation
of
mechanisms.
It's
all
everywhere
in
the
cell
has
an
equal
chance
of
this.
These
parameters
being
correct,
and
so
that's
why
they
say
well
stirred
and
then
a
reaction,
diffusion,
Master,
equation
description.
A
A
C
A
C
A
C
B
C
B
C
B
C
C
C
B
C
A
B
C
C
A
I,
don't
think
so
they
just
mentioned
that
things
are
either
well
mixed
or
not,
and
a
lot
of
this
might
be
because
you
have
you
know
hydrophobicity
of
DNA
and
at
that
scale
it's
like
that's
really
kind
of,
because
it's
pretty
packed
in
as
you
see
from
the
sea.
So
this
is.
This
is
the
cell.
This
is
the
model
of
the
cell
here,
so
we
have
this
Digital
model
of
the
cell.
This
is
the
image
or
we're
taking
these
coordinates
from
the
image
and
then
we're
building
this
model.
A
And
then
this
is
what
we're
simulating.
So
you
get
like
these
macromolecules
that
are
maybe
clustered
or
not
you're
moving
around
and
they
can
interact.
You
know
they
can
have
a
do
some
reaction
anywhere
in
the
in
the
cell,
which
is
of
course
very
small,
and
then
they
through
the
spatial
simulations
they
do
I,
don't
know
if
they
really
break
it
out.
I,
don't
know
if
they're,
comparing
the
two
well-mixed
and
spatial
that
might
be
further
down.
A
A
A
Gene
expression
between
will
stirred
and
spatially
resolved
simulations,
so
I
guess
the
spatially
resolved
is
where
they
have
the
coordinate
information
so
that
they
know
where
and
then
so.
They
just
have
accounts
for
RNA
mRNA
for
genes
coding
for
genetic
information
processing
proteins
over
time.
This
is
the
MRNA
count.
This
is
the
time,
so
you
can
see
that
there's
in
the
spatial
resolved.
B
A
A
Are
being
incorporated
into
mRNA
in
the
I
guess
in
well
stirred
model
and
then
second,
the
current
spatial
model
does
not
include
DNA
replication,
which
initiates
on
average
around
10
minutes
in
the
well
Stern
model.
So
this
is
yeah.
It
still
doesn't
really
give
a
lot
of
insight
into
that
difference.
So.
A
A
C
B
B
B
A
Set
is
the
concentration
of
intracellular
pools
of
NDP
ntp
and
ADP,
so
these
are
macromolecules,
and
you
can
see
that
there's
a
difference
here.
The
spatial
spatial
result
room
again
has
more
variation
and
fluctuation
the
scale
protein
counts.
These
are
a
number
of
unique
proteins.
There's
a
much
difference
there
and
then
the
number
of
unique
mrnas
are
the
this.
The
accounts
for
the
half-life
in
minutes.
A
So
this
is
just
things
that
are
getting
synthesized
and
degrading,
so
you
get
more
more
unique
mrnas
in
the
spatial
resolved
model,
but
it
decays
a
little
bit
faster.
At
least
you
know,
according
to
this
histogram
and
see
so
in
terms
of
our
mRNA
Half-Life
they're,
compared
between
the
two
methods,
where
the
well-stirred
half-lives
depend
on
the
act
of
the
greatest
sum
statistics
in
the
spatial
model.
So
this
is
something
that
there's
a
linkage
here
between
the
two.
A
The
well
stirred
model
is
longer
half-lives
on
average
than
the
half-lives
in
the
spatial
model,
as
we
see
here,
and
they
don't
really
see
why
that
is
exactly.
But
but
there
is
also
agreement
between
the
Wall,
Street
and
spatial
simulations
and
the
MRNA
comes
in.
The
spatial
model
are
higher,
in
average,
likely
due
to
the
difference
in
transcription
rates
between
the
two
models.
So
it's
the
spatial
model
also
has
lower
concentrations
the
nucleotides,
and
then
we
talked
about
that
so
yeah.
There's
there's
a
lot
of
like.
A
A
A
A
C
C
A
C
A
A
A
C
B
A
Yeah,
that's
okay,
yeah
well,
I
think
that
was
yeah.
That
was
a
really
good
paper
for
understanding
how
people
are
simulating
cells
and
having
the
parts
there
in
a
very
simple
cell
that
would
describe
sort
of
a
function
in
a
cell
and
what
you
need
there
and
how
those
things
behave.
So
I
mean
Indians,
larger
insights.
I
know.
The
whole
point
of
this
is
to
look.