►
From YouTube: Ken Ho Talk (RIKEN)
Description
Presentation to DevoWorm group, June 6th 2016. Ken Ho from RIKEN (Japan) on Biological Dynamics Markup Language (BDML) and Systems Science of Biological Dynamics (SSBD).
Paper: Kyoda, K., Tohsato, Y., Ho, K.H., and Onami, S. (2015). Biological Dynamics Markup Language (BDML): an open format for representing quantitative biological dynamics data. Bioinformatics, 31(7), 1044-1052. doi: 10.1093/bioinformatics/btu767.
A
Came
to
Japan
about
20
years
ago,
my
background
is
actually
on
AI
and
robotics
and
into
that
and
I've
shift
and
all
sorts
of
different
things
now,
I'm,
actually
working
on
bioinformatics,
really
our
system
or
system
biology.
That's
the
lab
that
we
are
basically
in
Japan.
Okay,
we
can
just
say
something
about
my
lap.
Rick
can
is
probably
one
of
the
largest
research
institute
in
Japan,
so
their
ample
funding,
the
place
that
we
are
in
actually
is
in
Kobe,
and
the
one
of
the
center,
which
is
more
famous
than
we
are,
is
the
weekend.
Cdb.
A
Or
cell
development
biology
center,
but
we
are
actually,
although
we
attached
to
it,
we're
within
it,
but
we
actually
belongs
to
another
center
core
cubic,
which
is
a
quantitative
biology
research
center.
So
our
our
working
mainly
on
developing
technology
on
like
computational
biology
simulations
looking
at
whether
we
can
assimilate
are
one
cell
kind
of
mathematically
or
simulate
numerically
on
so
develop,
not
lost
cell
development
in
itself,
but
on
cell
biology,
but
our
mind.
A
My
lap
that
I
that
I
involved
with
is
mainly
on
C
elegans
to
so
many
on
developmental
biology
and
on
C,
elegans,
embryogenesis,
etc.
Now
going
back
into
the
development
of
the
our
database,
so
I'm
just
prepare
presentation
asst
what
Richard
was
asking
try
to
see?
How
do
I,
when
I
get
into
the
sharing.
A
A
C
A
Black
doesn't
work,
ok,
wait
a
minute:
can
you
see
anything
on
now,
I
see
them.
Ok,
this
is
better.
Ok,
so
I
just
used
this
one.
Ok,
how
Apple
does
it,
though?
Ok
so
on
the
motivation,
I've
gotta,
so
basically
we
the
reason.
What
we
want
to
do
is
on
a
quantitative
just
to
get
numerical
data
from
from
images
for
something
like
this
one.
Where
are
you
going
so
there's
a
embryogenesis
of
C
elegans
and
all
those
direct
observation?
Is
that
Norm
looking
at
it
as
well?
A
Can
the
computer
see
it
basically
nowadays
with
with
image,
processing
and
computing
powers?
Now
we
can
actually
extract
different
parts
of
of
an
embryo
for
somebody
in
here
we're
actually
extracting
the
nucleus
of
the
embryo
and
in
in
three
dimension,
and
there
are
actually
a
lot
of
all
this
data
of
around.
This
is
zebra
fish
from
Keller
and
fruit
fly.
A
Show
you
later
on
more
Morrible,
but
at
the
moment
what
we
found
is
that
it's
very
difficult
to
to
use
the
old
estate
that
they
have
it
comes
in
all
different
formats.
It's
different
difficult
for
us
to
locate
them
really,
and
so
the
idea
of
of
this
is
actually
to
somehow
to
have
a
uniform,
unified
format
for
putting
in
all
this
data
and
also
to
have
a
database
that
allowed
us
to
do
comparisons
to
do
more
analysis
of
this
data.
A
Can
you
see
the
slider
or
transgendered
again?
It's
variable.
Ok,
Sokka
project,
it's
actually
funded
by
the
support
of
the
kind
of
Japan's
national
bow.
Our
science
database
end
it's
part
of
a
larger
project
on
LiveScience
database
integration
in
Japan,
so
they're,
basically
trying
to
get
more
databases
in
japan
on
bioscience
together
into
in
the
111
entity.
Some
sort
of
such
that
we
can
cut
people
can
search
as
well
as
can
do
more
with
it.
So
in
some
ways
it's
support
by
the
government
and
and
with
with
the
ministry.
A
Actually,
you
know,
basically,
all
the
researchers
in
Japan
have
an
application
to
support
the
project.
So
there's
a
lot
of
support
in
in
in
in
in
terms
of
getting
data
and
for
for
other
people
to
use
the
format
and
etc
that's
within
Japan,
and
so
we
are
hoping
that
we
can
actually
share
this
for
the
wider
world.
So
our
aim
is
basically,
as
the
funding
goes.
The
aim
is
to
store
and
share
the
or
this
or
this
different
type
of
biological
data
to
to
the
community,
and
basically
people
can
can
access
it
directly.
A
A
The
technical
question
is
actually
yes,
when
we
design
the
I'll
show
you
later
on,
for
example
like
the
pd
ml.
What
we've
done
what
happen?
Now
we
have
people
coming
to
us
and
saying
that
hey
we
just
we
don't
have
any,
but
how
does
a
dynamical
data?
So
it's
time
point?
Is
it
go
to
0
so
to
speak
or
time
scale
0
is
equal
to
0.
So
it's
a
bit
odd,
sometimes
that
things
happen
and
sometimes
example
like
in
our
format.
We
actually
have
organisms,
so
we
actually
state
some
sort
of
organisms.
A
But,
however,
some
of
the
simulations
that
some
of
our
colleagues
stars
they
don't
have
any
is
a
is
non-descriptive
organism
probably
can
be
human
can
be
any
mammals,
so
is
sometimes
can
we
can
get
into
that
house?
The
idea
is
that
we
can
extend
those
things
hopefully
and
can
adjust
to
it.
But,
yes,
we
do
find
from
time
to
time
different
requests
that
we
haven't
thought
of.
So
this.
A
No
okay,
so
I
carry
on
them,
okay,
so
just
so
what
we've
come
up
with
as
well
as
set
earlier
is
about
this.
Creating
a
basically
a
unified
language
in
us
in
or
what
I
would
say
is
a
file
format
that
that
we
can
actually
encode
or
this
types
different
data.
So
we
come
up
with
the
what
we
call
a
BD
ml
biological
dynamics,
markup
language,
it's
just
another
XML
languages.
A
It
basically
can
represent
five
different
types
of
objects,
be
a
point,
align
a
face,
a
circle
or
a
sphere,
and
you
can
then
define
it
within
it,
as
in
here
in
the
slide
with
we
were
able
to
actually
represent,
have
a
flexibility
of
represent
different
type
of
things,
from
molecules,
cell
divisions
to
C,
elegans
behavior,
to
nuclear
dynamics
and
accessory
sectors
in
different
scale
and
in
four
different
organisms
for
simulations
and
measurements,
etc.
So.
A
Basically,
we
put
them
all
into
a
database
called
SPD
and
it
is
open
access
to
everybody.
So
if
you
click
on
this
website,
you
can
you
can
get
to
it
and
now
I
should
do
a
demonstration
on
some
of
the
parts
of
it.
So
you
can
actually
get
some
of
the
data
sets
of
both
quantitative
data,
for
example,
numerical
data
that
we
actually
extract
it,
for
example,
from
from
image
processing
and
also
by
hand.
Quite
a
lot
of
those
things
are
actually,
although
done
extracted
automatically.
A
A
Okay,
don't
know,
any
of
you
are
interested
in
this.
Just
quickly.
Go
through
basically
is
internally
is
just
a
data
SQL
database
with
a
application
framework.
I
don't
know.
Are
you
aware
of
the
amuro
database
system,
which
we
can?
You
can
store
images?
We
actually
use
utilize
this.
This
is
also
an
open
source
software
developed
by
the
University
of
Dundee
in
the
UK
and
so
far,
I
think
this
is
more
interesting.
A
What
is
currently
available
on
SPD,
so
we
have
ecoline
single
molecules,
but
we
also
have
a
lot
of
C
elegans
nuclear
experiment
from
bow
and
Keota,
which
is
a
one
one
of
my
colleagues
he's
got
a
100
if
he
sets
of
RNAi
C
elegans
embryogenesis
data
up
to
they're,
using
the
circulators
actually
using
DIC.
So
it's
only
up
to
about
32
cell
divisions
and
also
we
have
others
from
toshema
and
things
like
that,
and
we
also
have
simulations
Kimura
actually
do
some
simulations
on
C
elegans
from
nucleus
microtubule
links
like
that
and.
A
A
A
A
B
A
A
A
A
A
A
You
can
actually
see
the
visualization
of
it,
and
this
is
actually
sorry.
I
click
it
too
far
too
far
anyway,
so
you
can
actually
scroll
it
and
to
see
how
is
split
and
how
how
the
nucleus
is
extracted
together,
and
you
can
also
see
in
actually
is
in
24
D.
So
you
can
actually
see
you
can
actually
zoom
in
and
see.
The
actual
data
sets
how
it
was
how
it
was
connected.
Now,
if
you
I
go
back
sorry
about
that,
I
go
back
and
before
the
summary
page,
you
can
actually
see
the
actual
image.
D
B
A
It's
all
right,
I'm
just
trying
to
say
you
can
actually
sorry
about
the
screen.
Size
is
some.
You
can
actually
see
actually
the
actual
EIC
image
data
set.
So
I'm
loading
all
the
data
set
this
all
available
for
you
to
download,
or
you
can
view
it
on
screen
or
whatever
so
and
as
you
can
see
how
each
slice
of
the
images
are
extracted.
A
So
if
I
click
on
this
one
I
don't
know
what
can
you
see
this
one
I,
exactly
what
you
will
be
able
to
see,
but
actually
you
can,
you
can
actually
see
you
can
actually
see
the
actual
microscope
microscopy
image
and
with
that
we
extract
each
each
of
them.
We
extract
us
a
line
as
an
outline
of
the
off
of
the
nucleus
into
a
line,
and
you
can
act
see
different
layered
how
how
how
it's
been
extracted
and
all
this
data
are
actually
becomes.
A
So
all
this
data.
Unfortunately,
this
visualization
have
not
have
not
include
the
annotation
data,
so
the
annotation
data
is
actually
available
on
on
the
data
set
and
on
top
of
it
it's
also
have
an
image
you
can
play
around
with
which
your
oven
actually,
rather
than
actually
get
into
displaying
that,
but
one
of
the
two
sexy
we
go
into
another
two
sets
understand
along
two
sets
with
the
BD
ml
viewer
actually
provide
you
with
some
some
of
the
lineage
theta
available.
B
A
B
Slice,
it
says:
half
a
micron.
Ok,
my
name
is
ours?
Serious?
Okay!
Oh
it's
half
a
micron,
okay,
okay,
which
is
about
the
limit
of
light
resolution.
Okay,
so
let's
run
something:
oh
yes,
sir,
if
the
DIC
data
corrected
for
blurring,
you
know
when
you,
when
you
do
like
microscopy,
your
vertical
blur
is
greater
than
the
horizontal
okay,
you
convolve
for
that
yep.
It
is.
A
B
A
B
A
A
Well,
yes,
we
can
even
access
at
the
moment
that
the
visualization
to
that's-
not
it's
not.
Of
course
you
can
I
mean
it's
just
data
and
I'll
show
you
later
on,
with
with
a
python
data
tools
that
you
can
actually
just
extract
whatever
part.
Okay,
okay,
so
I
mean
that
it's
just
data.
This
is
not
a
movie,
so
you
can
actually
this.
A
D
E
A
A
That's
that
DIC
image,
so
so
what
we
are
doing
is
just
we
are.
We
are
kind
of
extracting
the
outline
of
the
of
the
nucleus
and
with
that
sometimes
it's
not
actually
smooth,
that's
what
I
mean.
Sometimes
you
just
and
and
and
we
just
leave
it
as
that
I
say
a
line
rather
than
saying
that
have
a
definition
of
what
the
outline,
whether
it's
a
smooth
one
or
not,
a
smooth
10.
What
really
great.
A
B
A
A
A
And
this
data
set
is
195
time
for
time
point,
so
it's
much
larger
and,
as
you
can
see,
actually
give
you
more
because
the
killer
one
because
of
DIC
it
have
a
limit
on
of
of
all
the
nuclear
or
the
cell
division
that
they
can.
It
can
actually
record
so
this
one
actually
gives
you
a
more
complete
set
of
embryogenesis
of
of
C
elegans
and.
A
A
Yeah
except
okay.
So,
although,
although
although
although
they're
in
the
different
for
a
minyan
in
the
same
file
format,
the
data
can
be
all
different
and
yes,
those
are
different
data,
they
extract
it
differently.
They
do
different
things
on
it.
The
DIC,
because
this
is
life
is
not,
it
doesn't
have
a
GF
p,
so
you
can
say
that
the
wild-type
is
work
hard.
Well,
you
know
this
one,
okay,
so.
E
B
A
A
A
D
A
D
A
Hello,
it
is
still
okay.
My
computer
is
playing
up.
Okay,
sorry,
okay!
So
if
you
go
in
here
when
you,
when
you
see
that
the
images
you
can
actually
have
a
download
image,
download,
BD
ml,
so
the
distal
actual
data
file,
final
images,
you
can
download
the
images
zip
files
view
it.
You
just
seen
it
and
you
can
actually
view
the
images
here
separately.
So
just
quickly
about
you
can
download
the
images
now.
A
A
So
you
can
actually
go
through
a
restful
api
and
you
can
actually
example
in
using
an
API.
You
can
actually
use
different
computer
languages
or
hour
or
whatever
to
actually
get
the
data
directly
from
SPD.
So
you
don't
actually
need
to
download
all
of
the
data
sets
in
order
to
view
it.
You
can
actually
do
a
part
of
it.
The
API
is
given
in
our
website
here
you
can
have
a
look
at
it.
I
also
actually
put
in
a
Python.
A
A
E
A
At
the
moment
it
doesn't
it
actually
use
more
more
than
that
to
to
dip
the
browser
view,
but
I'm
not
just
showing
you
is
that
you
can
actually
do
a
similar
view,
not
as
not
as
interactive
with
with
just
the
API
okay.
So
sorry,
this
is
the
Python
API
to
get
into
database
I'm,
just
restart
and
I
just
continue.
I
just
continue
running
it.
A
So
I
just
load
all
this
modules
reloading
it
and
one
other
thing
that
could
display
3ds
user
called
ivy
show,
which
is
not
it's
a
bit
unusual,
but
it's
in
there,
so
you
can
even
tried
it
and
then
you
can
load
Jason
and
you
can
load
the
API
itself
and
then,
after
that,
you
just
be
just
fine
it
and
now,
by
using
this
one
I'm,
just
loading
assimilation
and
then
it
can
actually
display
the
result.
I
API
us
I
say
as
the
data
so
I'm
just
showing
you
the
result
data.
A
A
So
you
can
actually
see
that
all
the
data,
so
I
need
to
load
this
function.
This
is
actually
to
load
the
the
function
to
get
all
the
coordinates
at
a
step.
Specific
time
point
so
here
I
choose,
they
choose
a
time
coin.
So
I'm
loading
it.
So
you
can
actually
see
that
I
I
should
load
fifteen
time
point
15
and
those
are
data
set
XYZ
and
T
and
I
think
it's
the
bio
data
set
that
the
one
that
you
guys
also
have
download
and
by
doing
that,
I
can
do
them.
A
B
A
You
enjoy
a
line
on
to
it,
I
think
it
actually
come
with
I'm,
not
so
sure
about
a
bow
one.
The
Hilda
one
come
with
a
link
core
yeah,
it's
called
p.
Is
it
searches
through
for
you?
It
actually
come
with
annotation,
as
well
as
as
well
as
the
link
age
at
the
leak
at
the
lean
age,
certainly
nation,
the
tracking
data,
easily
okay,
okay,
so
some
of
the
data
set
come
with
the
tracking.
Later
some
of
them
doesn't
depends
on
the
beta
provider.
A
A
So
not
only
that
we
also
actually
in
our
website,
you
can
actually
download
a
image,
a
plug-in
tools
that
you
can.
Actually,
if
you
have
images
you
can
actually
do
your
own
PDM.
L
at
the
moment
is
still
the
saving
the
PDM
l
is
still
thing
is
still
under
to
be
released,
but
some
part
of
it
is
already
there
and
you
can
actually
see
visualization
there's
a
standard
plug
in
image
j
plug
in
visualization
to
to
look
at
to
look
at
the
pd
ml
file.
A
A
A
So
so,
some
of
the
things
just
you
can
play
around
with
okay,
what
happened
with
the
SPD's
that
we
also
open
source
the
software
for
saving
all
those
things,
so
we
actually
put
it
into
a
github
call
open
as
a
speedy
it.
The
only
thing
difference
is
that
it
doesn't
include
the
amuro
unity,
little
cow,
install
it
separately
and
to
link
it
separately.
A
The
differences,
basically
it
runs.
Spd
currently
is
run
on
Red
Hat
surfer.
Let's
open
SS
3d
run
on
ubuntu.
It
has
some
limitations
reading
different
different
version
of
pd
ml
files.
At
the
moment
it
doesn't
support
face
yet,
but
hopefully
we'll
we'll
put
it
on
soon.
It
also
does
not
support
multi-part,
a
BBM
air,
yet
browser
visualization,
the
visualization
they
use
that
I've
shown
you
on
the
browser.
At
the
moment,
it's
only
4
SPD,
because
I
actually
need
to
treat
it
manually
to
to
some.
A
Sometimes
it's
actually
at
the
lower
the
resolutions
in
order
to
get
a
smooth
beta
sets
to
2-under
on
the
screen,
whereas
for
this
one,
so
the
browser
visualization
is
only
on
single
time
point,
so
it's
just
to
make
it
easier.
I
also
have
a
new
read
file:
API
tu
tu
wien
files,
so
you
can
actually
also
play
around
with
it.
If
your
faders,
you
do
one
cup
create
your
own
database
and
you
can
use
both
co2
things
like
that.
Sorry
or
how
do
you
how
this
SPD
and
open
SPD
can
be
used
together?
A
What
we
are
seeing
is
one
of
data,
bizarre
Manas
data
factoids,
so
we
are
CAF
SSP
DB.
Hopefully
we
are
going
to
accumulate
and
store
or
craft
big
large
data
set,
genome-wide
or
genome
scale.
Data
sets
were
so
Vanessa.
Speedy
gave
me
more
research
groups.
Somebody
want
to
share
some
data;
they
can
do
their
own
now,
the
problem
at
the
moment.
We
don't
have
a
solution
yet.
But
what
is
how
we'll
get
into
the
same
problem
that
I
that
I
I
mentioned
before?
A
How
do
we
actually
be
able
to
search
all
the
data
across
across
the
internet?
You
know
so
to
speak.
One
of
the
solution
that
we
are
working
on,
poly
with
with
our
funding
agency,
is
using
ldf,
so
is
using
ldf
formatted
data
at
a
moment
that
RDF
data
of
SPD
is
actually
available
on
other
database.
What
so
out?
A
Ours
is
our
only
on
the
endpoint
so
on
the
weekend
database,
as
well
as
the
another
database
here,
so
you
can
actually
look
into
more
things
by
five
can't
we
elegans
you
can
be
able
to
come
more
interesting
things,
the
jeans,
etc,
etc.
So
we're
hopefully
using
our
DF
or
similar
technology.
We
can
link
different
data
sets
together.
A
A
Quantitative
data
is
still
difficult
to
get
people
to
do
quantitative
data.
As
far
as
to
give
us
the
all
of
the
data,
we
hopefully
can
add
more
there's
two
sets
which
are
I
mentioned
earlier,
which
of
May
of
interest
to
all
of
you,
which
hopefully
will
be
released
either
within
this
year
or
next
year,
which
I'm
sorry,
which
is
two
of
my
colleagues.
They
are
doing
it.
A
Actually,
she
actually
extract
thatõs
from
fino
Bank,
the
movies
or
finna
bangs
on
genomic
data,
and
then
now
it's
become
a
quantitative
data
and
she
did
some
analysis
on
it,
and
so
that
will
also
be
release
soon
in
the
field,
but
within
a
year
or
so
so
those
are
22,
interesting
ones.
Other
thing
I
forgot
to
mention,
which
quite
important
at
the
moment
we
have
a
policy
for
storing
bathers.
So
all
data
at
the
moment
are
have
to
be
published
in
peer-reviewed
journals.
A
However,
we're
starting
to
receive
requests
for
data
to
be
stored.
We
just
some
would
like
to
be
published
like
just
being
reviewed.
So
at
the
moment
the
policies
is,
we
only
accept
we
only
share,
or
we
only
store
and
share
data
that
is
published
in
peer-reviewed.
It's
just
to
ensure
that
we
have
the
quality
of
data.
A
Just
we
say
that
what
you
can,
if
you're
here
you
have
trouble
with
the
data,
you
can
actually
go,
go
back
to
the
to
the
to
the
to
the
journals
or
to
clarify
with
the
with
the
author,
but
maybe
we'll
change
this
in
the
future
because
of
the
requests
of
sort
of
pre
pre
published
data
sets
so,
but
at
the
moment
we're
still.
Basically
all
the
data
are
that
are
available
have
been
previewed
in
journals.
A
Okay,
so
next
part
of
what
we
are
doing
at
the
moment
is
we're
actually
expanding
it
a
bit
further
by
the
request
of
our
funding
agency
and
that
we
are
actually
also
going
to
store
large
microsemi
cross
free
images
that
have
not
been
processed.
So
we
may
not
actually
have
the
quantitative
data,
but
just
the
images
and-
and
mostly
we
are
likely
to
be
still
concentrate
on
images
that
are
dynamical
so,
for
example,
Genesis
or
or
cell
divisions,
or
something
instead
of
just
just
a
single
picture,
although
we
are
not
excluding
them.
A
So
if
you
have
a
super
resolution,
data
sets-
and
we
like
this
sword-
will
probably
store
it
at
the
moment.
So
we
are
going
to
expand
the
database
in
the
storing
more
images,
a
future
plan
of
it.
We
are
looking
into
changing
some
part
of
the
internal
system
from
SQL
database
through
HTF
or
five
probably,
and
we
are
also
likely
to
more
of
a
distributed
Street
architecture
for
or
the
SPD,
the
other
part
of
it
is
that
auto,
be
DML
include
ldf
and
oncology
in
the
database.
Actually,
its
itself
have
not
actually
include
ontology.
A
A
So
in
summary,
the
SPD
is
a
database
developed
and
funded
by
the
by
Japan,
a
mainly
supporting
Japanese
scientists
in
storing
and
reusing
those
database
at
data
sets,
but
also
we
hopefully
be
able
to
facilitate
and
contribute
to
others
too
slow
and
share
data,
and-
and
we
can
do
do
more
things
with
it.
So
here
you
are
just
the
end
of
my
talk.
Hopefully,
that's
not
too
long.
E
A
Store
everything
but
the
other
images
are
stored
in
a
separate
system.
An
architectural
e
is
stowing
on
one
server,
but
is
integrated
with
it
at
the
moment.
The
API
does
not
cover
that,
yet
we
are
looking
into
it
and
also
we
are
looking
into
overlaying,
for
example,
images
in
the
visualization
on
the
browser.
We
are
thinking
about
overlaying
the
images
on
top
of
the
of
the
quantitative
data
or
to
have
the
side
by
side
such
that
you
can
have
a
most
smooth
so
to
understand
how
the
images
related
to
the
quantitative
data
now.
E
E
A
B
A
Happens,
innocent
and
this
another
thing
is
quite
interesting.
We
have
disgusted,
but
we
haven't
actually
well
the
idea.
Okay,
a
lot
of
groups
are
asking
us
about.
Oh
can
I
store
the
images
and
you
automatically
inmates
processed
and
extract
those
things
for
us,
and
we
say
no,
and
so
at
the
moment
we
do
not
actually
have
the
capacity
to
to
do
the
image
processing
algorithm,
as
well
as
the
two
to
be
available
as
a
software
as
well
as
a
surface,
but
maybe
in
the
future,
don't
know,
but
I.
A
We
haven't
actually
terms
of
how
the
incorporate
that
we
haven't
actually
have
done
anything
on
it.
One
suggestion
is
actually
to
have
example:
each
data
set
will
have
some
sort
of
metadata
to
tell
you
about
how
the
data
is
being
extracted,
although
is
already
done
in
a
database.
I
mean
ur
in
on
the
publication
in
the
paper,
but
we
thought
that
maybe
you
can.
A
We
can
actually
do
a
bit
more
to
have
to
have
a
way
in
such
that
you
know
that's
what
you
mentioned
about
Google
AI
and
things
like
that,
whether
we
can
actually
use
deep
learning
or
something
like
that
to
to
do
more
on
beta
extraction,
and
things
like
that
by
the
moment.
Yeah,
I
think,
is
really
difficult
area
of
quite
challenging,
but
but
we
haven't
included
in
as
a
surface
in
what.
B
B
A
B
B
Okay,
now
what
happened
with
this
is
my
mother.
Who's.
An
artist
about
10
years
ago,
went
over
the
this
image
in
detail
and
she
closed
all
of
the
boundaries
that
were
open:
okay,
okay,
that
had
caps
in
them.
It's
also
removed
all
of
the
occasionally
to
be.
We
digitized
the
whole
thing,
so
she
she
she
filled
it
a
gap
since
you
moved
any
extraneous
point
binary
image.
Ok,
so
the
binary
image
goes.
A
A
A
B
B
But
what
the
morphic
genetic
furrow
is,
it's
a
set
of
where
the
cells
are
contracting
at
the
apical
ends
all
right,
so
that
creates
a
furrow
in
the
surface
and
on
the
firm.
The
furrow
moves
slowly
across
the
tissue
all
right
as
it
moves
behind
it.
The
ommatidia
start
forming
the
little
groups
of
cells
that
form
the
tiny
lenses
of
the
fly.
I.
Ok,.
A
B
So
I
did
a
a
very
crude
analysis
with
my
wife,
Emily
Gordon
and
with
Hillary
alto,
where
put
one
lying
down
the
morphogen,
a
tackler,
oh
and
then
let
count
measured
the
distance,
the
intersections
of
that
line
with
the
cells,
uh-huh,
ok
and
then
we
did
what's
called
a
variogram
analysis
and
they
showed
that
along
the
north
of
genetic
borough,
you
had
two
large
cells,
too
small
cells
too
large
to
small
along
the
length
of
it.
So
this
detected.
B
A
B
Bradley
is
doing
further
analyses
where
we're
getting
the
bet.
The
number
of
cells
we're
looking
at
patterns
of
the
cells
and
various
kinds
of
distributions
and
we're
analyzing,
these
nine
thousand
cells
to
death
and
everything
we
can
think
of
okay,
I
and
what
you
see
is
a
progression
of
the
evolution,
the
progression
of
the
formation
of
the
the
groups
of
cells
that
later
form
the
lenses.
Okay,
ok,
now
is
this
kind
of
data
set,
something
that
would
go
into
your
yeah.
A
We
will
be
interested
in
data
sets
that
any
data
sets
will
be
happy
to
be
included,
but
what
we
are
more
in
what
we
that's,
what
we
said
the
pulse
at
the
moment
is,
if
you
have
a
peer-reviewed
journal,
that
that
described,
that
will
be
will
be
happy,
oh
well,
to
work
with
you
change
it.
We
actually
at
the
moment.
We
actually
do
the
the
conversion
of
your
data
sets
in
the
BDM
ourselves.
B
A
A
B
B
Lot
and
then
the
the
analysis,
so
you
know
we're
trying
to
analyze
it
and
see
what
interesting
things
come
out
of
it
now.
Our
hope
is
that,
once
it's
done
that
this
is
just,
of
course,
one
image.
One
time
point:
okay,
basically,
okay,
you
know
it's
it's
like
having
one
snapshot
of
the
development
of
the
station
okay,
but
we
hope
that
by
doing
this
analysis
we
can
inspire
people
to
start
getting
some
time
lapse:
data,
okay,
okay,
okay
and
if
you've
got
drosophila.
Colleagues,
maybe
they'd
be
the
right
people
to
do
it.
Yeah.
A
B
B
B
B
Okay,
while
it
turns
itself
right
side
up
again:
okay,
okay
and
then
this
will
give
us
a
set
of
images
of
the
surface
and
then
the
robot
rotates
the
embryo
and
flips
it
another
way,
and
then
it
rotates
another
way.
So
the
idea
is
to
get
images
of
the
whole
surface,
all
right:
okay,
okay-
and
this
would
be
on
a
live
embryo
and
we
would
do
it
in
time,
lapse,
yeah
and
then
get
the
whole
surface
of
the
embryo
over
time
from
the
single
cell
stage
through
at
least
neural
tube
closure.
Look.
B
E
D
A
B
B
The
segmentation,
no,
we
have
to
see
we've
got
a
two-step
thing.
First
of
all,
we
have
to
get
the
robot
to
work
the
embryos
to
cooperate
and
get
all
the
pictures.
Then
they
have
to
be
montage
into
a
four
dimensional
data
set:
ok,
ok
on
the
certain,
so
this
is
the
surface
of
the
embryo
versus
time
they're
going
further.
Then
you
do
the
segmentation
and
then
we
do
sell
tracking
and
also
what
we've
observed
on
these
embryos
or
waves
of
contraction
and
expansion
of
these
cells
uh-huh.
B
A
B
B
We
are
waiting
to
do
the
last
proofreading,
ok,
ok,
yeah,
and
so
should
it
will
be
printed,
probably
in
July
or
August.
Ok,
interesting,
ok,
yep
I.
So
this
gives
a
theory
of
it.
The
basic
idea
is
that
the
we
think
that
waves
of
contraction
or
expansion
of
the
cells
are
the
triggers
from
cell
differentiation.
Okay,.
A
B
Well
that
they
would
be
calcium
ways
that
have
a
calcium
component,
okay
and
that's
actually
lional.
You
know
the
work
of
land
on
Jaffe,
no,
no
okay,
I
know
happy.
He
died
a
couple
years
pick
studying
calcium
waves,
yeah
and
he's
published
a
little
bit
about
the
waves
we
observed
and
he's
predicted
that
they
are
calcium
waves
out
there
they're
very
slow.
He
calls
them
ultra
slow,
ultra,
slow
and.
B
B
A
D
A
C
A
Differentiation
is
already
half
set,
so
the
idea
of
hours
that,
where
we
can
actually
from
time
division,
we
can
actually
analyze
for
some
other
other
species
in
terms
of
when
the
day
actually
decide.
I
gotta
get
the
faint.
When
does
the
fate
decide
and
can
we
can
we
tell
can
be
these?
Can
we
tell
from
different
I'm
divisions?
That's
one
part
of
it.
A
The
other
one
is
person
that
means
more
interested
in
the
physiological
side
of
it
is
when,
when
the
cell
divided,
how
much
force
does
it
required
and
how
does
it
change
and
why
does
it
spin
around
and
things
I've
and
and
with
different
species?
You
can
actually
see
different
things
coming
out,
so
yeah
I
mean
it
is
not
easy
at
the
moment.
A
It's
not,
and
it's
not
obvious
either
how
we
can
you
use
the
data
in
in
different
species
to
analyze
certain
things,
but
I
think
because
of
the
data
is
now
available
together
in
one
place,
it
actually
give
you
a
possibility
of
actually
doing
some
analysis
on
it
where's
before
quite
difficult,
so
it
can
be
quite
interesting.
I
think,
but
I
don't
have
a
concrete
kind
of
example.
That
I
can
give
you
that
you
know
how
what
we
can
achieve
out
of
this
okay.
D
A
C
A
D
C
E
E
A
A
A
A
A
A
C
B
A
Do
have
some
tough
goes
on
in
terms
of
trying
to
do
more
understanding
and
doing
more
of
data-driven
of
biology.
So
is
how
do
you
have
all
this
data
and
how
these,
as
what
you
mentioned,
about
deep
learning
or
not
using
the
blurring
but
other
other
techniques,
in
order
to
look
at
cell
biology
and
as
well
as
in
embryogenesis?
A
A
Our
lab
is
yes,
it's
also
similar
in
terms
of
looking
into
using
C
elegans
as
a
basis
as
a
model
to
look
at
understanding,
different
genes
and
how
how
fat
and
basic
biology,
basically
so
we
are
doing
actually
kilda
is
actually
doing
what
what
is
trying
to
do
is
to
do
to
do
the
genome-wide
RNAi
knock
knock
out
as
well
as
a
phenotype.
So
now
we
are
creating
all
this
images
or
this
numerical
data,
so
it
can
actually
match
the
two
together,
so
you
can
actually
do
a
feature.
Map
of
of
different
gene
causes.
A
A
A
A
B
A
C
D
A
Mother,
you
can
the
nucleus,
but
it's
that
idea.
How
do
you
say
it?
It's
that
real
or
whatever
is
a
mocker
okay.
Well,
isn't?
This
is
for
real
I'm
trying
to
say
is:
we
must
gather
parts
yeah,
yeah,
yeah
yeah,
and
it
doesn't
it
that's.
Another
problem
that
we
are
working,
as
some
of
my
colleagues
are
working
on,
is
to
have
two
or
three
markers
in
order
to
get
the
membrane
as
well
as
a
nucleus,
etc.
At
a
moment,
do
you
only?
Can
you
know
you
can't
I.
C
B
A
B
B
A
B
A
A
B
A
B
A
B
B
E
B
B
Problem
dealing
with
fossils,
you
know
so
synchrotron
radiation
was
used
for
Reconstruction
era,
producing
very
nice
images.
Of
course
there
they
were
fossilized,
so
you
had
different
contrast
mechanisms,
uh-huh
I,
don't
I
could
I
could
check.
What's
going
on
on,
saw
soft
x-rays
and
embryos
and
see
okay.
C
B
Image
it
live
use
using
an
intense
x-ray
source,
it's
monochromatic,
okay,
okay
and
with
being
monochromatic.
It
means
that
the
absorptions
you
don't
need
any
of
what's
called
beam.
Hardening
Corrections.
If
you
have
a
ordinary
x-ray
tubes,
produce
a
broad
spectrum
and
it
complicates
the
analysis,
nice,
a
thinker
tron
you
can
get
very
monochromatic
radiation,
which
means
you
don't
have
any
of
those
problems,
so
it
r
and
the
software's
were
readily
available
for
reconstruction.
Well,
okay,
okay,
I
I
worked
on
that
kind
of
software
for
decades.
Well,.