Okay, we're live. Take it away.
Hi. Hi there.
Can they see us?
Yes, I guess so. Let's see whether they can...
You are live. You're live.
We are live. Okay, so here we are.
Hi everyone. Hi. I'm Noa.
This is Lawrence.
Nice to meet you. We are here
streaming from the Weissman Institute of Science
in Israel. And
together we are going to present
the
comparison between
cybersecurity and biology.
I hope you are ready.
Well, I'm ready, so I guess the stage is yours, Noa.
Okay.
So, nice to
meet you. I will just
present...
Okay. Sorry.
Okay, so our agenda
for today... I have two
targets with my presentation.
First of all, I'll show you how beautiful and complex
biology is. And the second
part will be to make you jealous.
So, who am I?
My name is Noa. I'm a master
student at the Weissman Institute of Science.
I'm a junior scientist in
a lab
that is doing research
specifically in microbiome, but also diabetes.
And I'm a former
cybersecurity researcher at
Checkpoint.
So,
until a year ago,
I was a
reverser in Checkpoint. My job
was to reverse...
I'm sure that you all know about this job,
to reverse engineering malware,
and then to implement
the defensive
coding into Checkpoint
antivirus.
So, you know, I thought
I'm a pretty smart girl, after all.
I even did
some fuzzing stuff. I found some
vulnerabilities.
You know, I
built myself up as a
hacker, as a researcher
in Checkpoint for many years.
And then...
But I was always
very curious about complex
themes, not only malwares, but also
about biology.
And a year ago, I decided
to come here and to become
a researcher in biology,
specifically about
the new area in
microbiome.
So I thought, you know, I'm so cool,
and I'm doing reverse engineering
for so many years.
So, you know, all of the
bacteria and the COVID-19 and whatever
comes up,
all of them
are going to be
scared of me, because now
there is a good reverser and a good hacker
that is coming into the field.
But then I found out that
apparently it's not as...
I think that we all...
I think that it will be really
easy for me to prove my point that it's
not as easy as we can...
as we thought,
until not very far.
You can
see, like, what happened when
just one bacteria just
did some
evolution and become
just a little bit
more dangerous than before.
Like, the COVID
is doing evolution
for many years and
like, I think it's really
easy to see how does it affect
the world right now.
And also
how I learned in a hard way
biology is way more
complex than any malware that I ever did
a reverse engineering in. This is
actually one...
just one slide of my
class. What you see here is
the complex system
of what happens inside
a cell when there is one
ligand that is binding
one of the receptors.
So, apparently
it wasn't the best idea for me.
I'm joking, of course.
I'm very curious and
this area is very challenging
for me.
And I think
that what we do here is just
beautiful.
And also
one more thing that I want to
point out during my lecture is that
when I tell people about biology
they all think that I'm
palpating and growing
microbes on a plate
but actually these days
computational biology is
on the rise and what we are doing
especially in Weizmann
is trying to take
each of the biology systems that we have
in our body or in bacteria or anything else
and build a model about it. This is
also one of my classes and this is from
Uri Alonlav that we are
doing a collaboration with.
So what he did here, he took
the aldenal gland
and tried to build
actual equations
I think I did something
okay, and tried to build
equations in order to
understand that this
thing here is the
complex system of
the glands and this one
is some
equations that are
explaining why
does diabetes type 2
develop
in so many people these days.
And now I will pass
the microphone to my friend that will
explain to you in detail about
some of the basic creatures that all of us
have in our body.
Thank you very much indeed Noor.
By the way, I just
want to point out that
the person
that actually introduced me
into this world and made me
very stressed
also a very
happy student but also a very stressed and a very
poor student is this guy.
So when your mom tells you not to talk
to strangers, maybe she has a point.
Okay, so I
guess I don't need any further ado.
Well, if you can see the camera, you will see that
in contrast to Noor, I will be sitting. I'm a little older
so forgive me for that.
So actually
when we were preparing this lecture
we
we
we came up
first with this title
Back to the Future
and me being this
very modest, humble
German dude that I am. I didn't
have all these fancy animations
that Noor showed to you.
But actually when I envisioned this presentation
I started with something
lame looking like this.
So again,
we are from the Advisory Institute of Science
in Israel and
what
I want to show you now
in the next 25 minutes
is in essence
how computer science
and biology not only go
side by side exhibiting stunning parallels
but also that they in fact
have an intersection that we can
exploit nowadays to address
current challenges that we
are facing.
But before I go into this
I actually want to fulfill
another request that Noor
kindly asked me which is
to explain a very central
concept in biology which by
itself also has
a counterpart
in computer science
and I'm talking about genetic code.
Okay?
Genetic code, our DNA
is basically the language
in which life is written.
So that's sort of
these chemical molecules that make up
four letters
A, T, C and G.
They write a
blueprint of life
but as such
they are not very useful because they have to be
transcribed. They have to be transcribed
from this DNA form, this genetic
information into just another
sort of code
which is RNA which then
is sort of translated
so put into
proteins, protein
molecules that make up all our cells
and therefore our body
as a central component.
And speaking about code, I think
again, it's not very far
fetched to reach from
biology, so the science
of life
to computer science because
you as computer scientists, mostly
out there in the audience, you will know what
a good computer
code is.
And basically I want to start
by asking the question whether
there is another analogy between
computer science and
biology. Are there any
parallels in terms of computer
viruses and biological
viruses because, I mean,
as you know, currently we are facing
a global
COVID-19
pandemic that originates from the
deadly spread of the
coronavirus, SARS-CoV-2
and I
want to raise the question whether
there's anything, if you
only know computer viruses because you're
a cyber expert, that you can
also learn about biological
viruses such as this
coronavirus.
So let's start with a computer virus, right?
With a computer virus that you probably heard about
the Melissa back in 1999
which
based on
a sequence like a stretch of code that you
can see on the left here
which basically with
some letters just gives
also the blueprint of a
malware basically that causes
a lot of damage and
I'm not sure whether you're surprised or not but what you're
going to see on the right now here
are the first 3,569
letters
of the SARS-CoV-2
this coronavirus, the version
Wuhan-U1
the genome, the genetic
information of this
virus which in total is about
30,000 letters long
that makes up all
this virus that now
spreads on the face of the Earth
and I'm
saying genetic information here but to be
precise this is not
DNA that you're seeing but
actually it's an RNA virus
so while this is
sort of the genetic information of
this virus, it's not
the same genetic information
that we have in our body sense but it's
only this transcript that I
introduced to you before.
But looking at this
sequence or this source code
of the virus, you wouldn't be
able to judge that this is particularly harmful.
Maybe on the left you could,
Melissa, because you're
an expert and you see what
it actually does
but there is no chance
even the best molecular
biologist looking at
the code on the right would be able to
tell, okay, this is a deadly virus.
But still, we
desperately needed to know this sequence
to detect this virus
but then the question
is always having this sequence, having
this source code, what does it do?
What is this sequence translated
to? And starting again
with the example
of Melissa,
basically what this virus did
is it prompted
always to send
emails to your own
Microsoft Outlook inbox or
started new
Word files which basically then
overloaded
and completely flooded
servers, like online servers and
stuff. So you had,
of course, I only made this up like
an infinite inbox
and
this is sort of how
the source code of the computer virus
translates into
a 3D structure, if you want to say so,
with a malfunction, so with a bad function.
And actually the same thing
biologists did then, they asked the question
okay, given the sequence of the
coronavirus, how does this
translate into what I mentioned before
these proteins, right, that make
up the shell of the
virus and
these different parts of the virus that
help the virus to infect
our body. And this is
basically one part of this
virus, so it's not the entire virus, it's just
one tool, actually what I said
like one element, one
part of a molecule that helps this virus
to infect our body, our
body cells.
And this is how this virus does
harm, right, it infects cells and then
eventually kills them.
So this is how in biology
a sequence or a source code
translates to a structure
and then a certain function.
And then to
sort of visualize how it looks
like in total, apart from
what it does, the
Melissa virus to overload
your inbox and the coronavirus
to infect your body cells,
actually
for the Melissa virus then it took
an artist to visualize
how this overload looks like
and this is how an artist apparently pictures
how a computer virus
looks like, which as you
can probably better judge than I do
is just
more or less arbitrary
representation.
But in fact, in biology, again, we can
use this wonderful
invention of microscopes to
actually look at how the virus looks
like. However, conventional
light microscopes are not sufficient enough
here, so actually we need electron
microscopes. However, if we use
these very sophisticated
apparatuses that cost like
a million dollars each and
require a lot of expertise to use them,
then we can actually see a virus inside
of cells.
So all these blue particles
that in this Forbes article
are described as
pepperoni, sorry,
slices of sausage
on your pizza, they are
the actual virus. However, I also must
say the truth is
if you use an electron microscope to observe
the coronavirus
in cells,
so this is actually the coronavirus of
the first identified
US patient
that was imaged here,
this original picture from
the electron microscope had no color.
So this blue
color is just put up
artificially in the end to visualize
the virus particles.
But here
we are still talking about
only
the size of
a meter divided by
one billion. So it's
in this order of magnitude.
So it's very small.
And you need sophisticated,
as I said, microscopes to actually see it,
but you can see it. You can understand
how this sequence source
code information that we saw,
how this translates to
first individual
aspects of this virus,
individual molecules, and then the virus as such.
Now, once it
has done its terrible job infecting
our body cells.
And this sort of
shows that
to some extent there are
parallels between computer science and
biology, in this case computer
viruses and
biological viruses such as
SARS-CoV-2.
But
is there, you know, besides
from
this example,
is there something we can
do about this? Is there something we can use
from computer science
to help us
fighting this
terrible pandemic that we are facing
in the US and all over the globe?
Is there an intersection between computer
science and biological aspects of
this pandemic?
This is a question Noah and I basically
asked ourselves. And in the very
beginning of this pandemic
in March, in fact, we
participated in
a hackathon that was
initiated by
InDev in collaboration
with Microsoft. They called it
HackCorona.
Creativity will not be quarantined.
And in the next
five to ten minutes, I would like
to introduce to you this
project that we initiated there to show
you how computer science can actually facilitate
in a very
concrete example, our
fight
against this virus.
And the
question that
we asked basically is
how
can we use computers to
combat COVID-19?
And the particular question we were wondering
about, because this was of major
concern in the beginning of this pandemic,
is how to
help people who develop severe
complications. Because we
understood that asymptomatic patients,
of course, they are very dangerous in
terms of unobserved
spreaders of this disease.
But in case
someone develops severe complications,
needs ventilation in the hospital,
then this
might
present a major challenge for
the health system in Israel
and all over the world.
So how can we basically assess
the uncertainty whether a certain
COVID-19 patient will develop
severe
symptoms and require
hospitalization and ventilation?
This is
the question we asked. And as you can see
in this picture, again, we were
pretty much using
biology as well as computer science
by these two mice that you can see here,
like a computer mouse and an actual mouse,
or at least a stuffed animal of a mouse.
And so, in simple
terms, our question was,
given the fact that we have
a certain number of
COVID-19 patients
that require, because
they develop severe complications,
hospital beds,
how to deal with them if the number of
hospital beds is actually
lower than the number of patients who require
hospitalization? In other
words, five
exceeds three, right?
So how to decide
which of them to allocate
to the hospital?
How to estimate what's the fraction of
people who develop severe complications
and how to decide whom to offer
with a certain priority these
available resources?
This was the question.
And the reason why we deal with this
was because at the Weizmann Institute
we are really at the forefront
of this, not only that we have
medical doctors on campus
who see these patients on a day-to-day
basis, but also the labs
we're working in from
Prof. Dr. Dr. Eran
Elinav and Prof. Dr. Ido Amit,
they develop now
technologies to not only
detect these
infections and stratify
these patients, but also streamline this
process, really linking
the foundational research with
the application in the clinics.
Okay?
So this is why we also felt
dedicated to address this
with our expertise, which
comes more from the
computational side here.
So what we did in the beginning
is, first of all, we asked, okay,
what's the data we are dealing with?
Because only if you have data,
you can actually use computational routines
to address
all these issues.
So we
basically simulated
based on existing
data,
an Israeli
dataset comprised of 10,000
cases, where we really used
Israeli demographic
information and health record data
that was publicly available
to a certain extent,
and set up
sort of a database where
we can get an idea, okay, this is Israeli
population, and now
given all this
demographic information and health record information,
can we basically now
estimate what's the fraction of people who
might develop severe COVID-19
complications? So first of all,
as you can see, we are struggling entering all this
data, but then what we obtained
is basically a more or less realistic
Israeli age distribution
here, and
also what seems to be
important, and we will hear about
diabetes and the metabolic syndrome later
on in Noah's part,
the distribution of the body mass
index in the Israeli
population,
which is basically a measure for your
body density,
so it's your body size divided
by your
body height to the
square, and it gives you an estimate
of...
if you have a high BMI,
you are considered obese, and then actually
this also affects your risk of
COVID-19 severe
complications, but also whether you get infected
in the first place. But it's not fully
understood, it's just an observation, so
anyways, we thought we need to include
this information, so first of all, we needed to have
some sort of realistic distribution
of this kind of
data. And then
what we were basically sampling from this information
we had back then when Israel
was not really facing many
many cases at the time
was this sort of distribution
of people
who got infected and then
sort of were
surviving through the disease, and
people who eventually died, and then you can
see that even though the overall frequency
of dead people, of course, or of dying
people is with the
infection or upon infection,
not necessarily caused by the infection,
is very low, but you can see that
if you're older, the
probability that you actually die
after getting infected with SARS-CoV-2
is higher.
So what we then
did is we used this data set,
this information that we
compiled, and
trained the model to basically
predict, given
a certain patient with
this health record and demographic
information,
what's the risk of developing
severe complications, and
then what you can always do is you can compare
your prediction to the actually observed
risk, because you have
information on this, right?
And then we get a pretty good agreement between
the predicted risk and the observed
risk here in this
preliminary data set.
And then we can ask the question, okay, what kind of information
is actually necessary to
predict this risk in a
reliable manner? And then
just as a test case,
proof of principle, we said,
okay, age seems to be
relevant and
critical,
so what happens if
we are
leaving out the age information? And what you
are seeing here is basically a receiver
operating characteristic curve,
a rock plot,
and basically if your
line, so your specificity
and sensitivity
of this prediction performs
pretty much randomly,
then this line that you see is close
to the diagonal line. So as you can see,
the orange line, if we leave out the age information,
we cannot really predict
that it's like randomly drawing,
okay, this person will develop
complications or this person
won't, but if we just consider
the age information as well in addition
to all the other information, then
this prediction becomes
very sensitive and very specific.
So it means that if we
say a person does
develop complications, then
the chance that this is a true
positive indication is
high. But also
if we say this person won't develop complications,
then also the chances
that this is a true negative, so it
indeed won't develop complications,
it's also sort of high.
So of course the perfect prediction
would be just going straight up
and then straight to the right,
but also in this toy model
or preliminary thing,
the performance of this prediction
was
not so bad, I would say.
So the way we
envision this basically is
once people get sick or
are suspected of being infected,
they have to see a doctor
who then puts in
all the information into a certain
database and some
device that collects
this information and then basically trains
this model that is based
on machine learning algorithms
and all these health record
information and then basically puts out
back to the doctor
a certain risk score.
So the application
tells the doctor, okay, this person
exhibits a certain probability to develop
severe complications, so better save a hospital
bed for this one, or in the ideal case
this person can just be
isolated at home, there is
a very low chance that it's going to be severe,
just monitor it carefully,
but most likely
the hospital bed allocated to this
patient can be
given to somebody else with a higher risk.
So, wait, what are we
lacking here? I already
presented to you that this machine learning algorithm
sort of works, but we need
this device, right? We need this application.
And this is also what we
sort of developed, like
something that can work on your mobile phone
and helps medical doctors, again, that
we have in our group and we talk to them on a daily
basis, to just use
it in a very handy way. If they
see the patient, they can check
in all this information in less than a
minute, and
then basically we can
just let
the machine learning algorithm do
its job and see whether there's
a certain risk.
And, of course, this
machine learning algorithm that I presented
to you, this model that we used there to make
the predictions, was just the first attempt
that we had in the Sakaton, you know, very
coarse-grained
first rough trial.
Needless to
say, you can also use
more sophisticated methods
where you can use all these
different
biological information on these patients
to assign how
important is this
information
to either say that there is a high
risk, or maybe it's just informative to say that there's
a low risk, or maybe a combination of both,
right? So we can, of course, you
always want more data. We can try
different models, and as
I said, we are in day-to-day contact with the clinicians
actually in charge, actually doing this, actually
doing the anamnesis or talking to the
patients,
suspected infected people.
So
this is what we envisioned,
you know, we can't really
not necessarily change
the number of people who are
susceptible and eventually will get
infected. It's just a matter of time,
it seems like.
And maybe
also, you know, the health system is sort of
currently at its limit, so we cannot
really expand the number of hospital beds.
But this app
that we developed, that by the way we
called Virologens, because it's like an
intelligent solution to virology
and inference of
this
virus-induced
health complications,
we could now use to assign
again hospital
beds to people who need it the most.
So in the scope of this hackathon, which of course is
very preliminary, and we are still sort of
thinking of how to
put this further,
in the scope of this hackathon, this
question was solved.
But as I already indicated, in the scope of this hackathon,
and this is just COVID-19, which
at the moment seems to be a major part of our
lives, but certainly it's not the only part
of our life, so
can we
make a more general
observation when we look at
computer science and biology?
Or in
other words, how can we
combine powerful
algorithms from computer science
with biological data sets?
And this is what I've shown,
what I'm showing you here, it's basically
the intersection of biological
data, this is how biological data
nowadays sort of looks like.
How can
we combine this biology with
this more like a conceptual or
theoretical view of
biology as a system?
As a living
system that just operates
certain commands and has
some input-output functions.
And the answer here is also the computer,
because only the computer will help us to process
this data and to put it into these
sort of models in terms of
understanding how biological systems
work. And this is a discipline
that I actually studied, I did my
PhD in this field,
which is called systems biology.
And let me
introduce to you the field of systems
biology with one of the godfathers of systems
biology, a guy who actually does
also a lot of robotics and
intelligent design and stuff, a Japanese
guy who also invented
robots that help people
to rescue
people from
collapsed buildings
after earthquakes and stuff.
Hiroaki
Kitano is his name, and
he describes systems biology as the
integration of experimental and
computational approaches to precisely
characterize biological systems
with quantitative data.
Okay, so now we have
quantitative data, which is machine-readable
from biology,
that we will use
to understand, in combination
with experiments in the laboratory,
how this
biological system works, simulating it at the
computer, using algorithms to
understand what's going on
there. And what does
go on there? How does this look like?
Well, you can imagine
as this sort of simple diagram
here, where basically on the
x-axis, you will have
time, and on the y-axis,
you will have the amount, quantitative
data, okay, of a biological
entity. Now,
what does this mean? So basically,
what you have in blue, these blue dots, these are
your data points, and then
with the computer, you can basically simulate the behavior
of the system, and you have some mean
simulation here, the blue line,
but then also you have this shaded region, which is
basically your uncertainty, okay,
your confidence interval. So not
only you get an understanding with the computer,
how does the system change over time,
how do biological entities
change over time,
but also, how certain are
you that this happens, right? The virus can
spread, like,
enormously, but maybe there's
also a chance it doesn't spread, or maybe there's a chance
that it spreads, like, even more.
And this is something we can quantify,
we can calculate. Sometimes
we have a problem with communicating this, but
there are possibilities in mathematics
to calculate these confidence
intervals and these uncertainties.
The way
this is done, the way we describe how
a biological entity changes over time
is, in mathematics,
called ordinary differential
equation. So this basically
describes to you how your biological entity
B is changing,
so the D represents the change, okay, it's the
change of B over
the change of time T, okay?
How does B change with time?
And this is now depending on
certain parameters, P, P1
and P2. And the plus
always means, you know, you get something more,
so depending on this parameter,
P1, you get more of B if you
already have B. So
if you have
something that can sort of reproduce,
like a virus can reproduce
or we can reproduce, you know, you
will get even more if you have more people to start
with, right? And then basically
the reproduction rate is your parameter
P, and it's positive because you get
more. But then
this curve of B is not only going up, it's not
only just getting more with time, right? It's also
going down. So there's also
another parameter, P2, and there's a minus
because it's getting less, that
tells me that
B is... the change of B over time
is getting negative. There's something
subtracted from it.
And of course, the more B we have, the more likely
it is that, you know, it destructs itself. You know,
humanity is very good at destructing itself
somehow, unfortunately.
But just by itself, it won't go just
up and down there. We need
another component here that influences this.
And this other component is A.
So for A, we can actually
write down a very similar equation.
So how does A change over time? Well,
it grows, but not only depending on A
but also on B, because, you know,
it destructs it, it eats it, for instance.
You know, humans are very good at eating
stuff, and also other living
beings.
But everything that lives also has to
die. So the more A we
have, the more likely it is that it's also dying
with another parameter, P4.
And then, you know, we get
more B, but then A gets more
to eat, but then eats more, and then
B goes down again. So
there's little A that can eat up B
and B can re-arrive, and then it's going up and down.
So this is sort of how nowadays
coupled ordinary differential equations,
because there's two of them and they are depending on one
another, we can describe the system's
behavior and even these weird
oscillations, like going up and down.
Also from a mathematical
point of view, a very interesting
observation of this system behavior.
But now
I talked about A and B a lot, but
saying B, I don't
necessarily mean a bee. Of course, a bee is also
a biological entity, right? The bee that gives
you honey and stuff. But B could be
anything, honestly. We are talking about
the human animal, the marine animal
like mice. It could be the number
of cells that are changing in our
body. It could be bacteria
that are also in our body, also cells.
It can be this virus, right? This is a
picture of a virus. But it
can also be just certain molecules that
change over time. For instance, here an antibody
that is produced against
this virus.
So
these are all options of biological
entities that we can describe in biological
systems. But to describe them
actually we have to think
on which level are we
talking.
Because biology happens on
many different scales, in space
and in time, right? The human being is
on the order of
magnitude
like
one to two meters, let's say.
Maybe basketball
players are more like two meters
and other people more like one meter.
Never mind.
But then we talked about
a mouse that is certainly not a meter
but more like centimeters.
And then we have cells that you have
seen that are more like micrometer
which is a meter divided by a million.
And then
in these cells we have small
subunits
that are even
smaller. Bacteria are also smaller
than our body cells.
And here we are then in the order of magnitude
of
a meter divided by
a billion, okay? So ten to the minus
nine, a nanometer.
so... and
all these scales, you know, changes are happening all the
time. That's how we are being alive.
Because molecules are turned over, cells
are turned over, our bodies are being turned
over. Everything happens there.
And this not only happens
on a very slow scale, like billions
3.6 billion years, this earth
does exist and life
does exist.
But also in a matter of hours,
in a matter of days, in a matter of minutes, in a
matter of seconds, you know, molecules
change their conformation in the order
of microseconds. Again, a
second divided by one million.
And this all, you know,
this is all part of this one
biological system that we are trying to describe
and there's no one out there who can
tell me that there's a way to
handle this complexity without a computer,
okay? So we need computer science
to address all this, to understand
biology at all these scales.
And if you're curious
for this, don't you worry
because there's a wonderful database at
bionumbers.org initiated
at Harvard Medical School
with you guys in the US
where you can ask the question, okay, how big is a cell?
How many molecules are in a cell?
How does a human body cell
is different from a bacterial body cell,
from a fruit fly, from a
worm? All these numbers are there
with references. You can look up the
original scientific literature to see
the evidence for these numbers. And it's just
wonderful looking at
all this information that
is out there. And not
only this information is out there,
but it's being continuously generated.
We are living in a time now, in an
era where biological data is
not, for the first time, not a bottleneck
anymore. You don't need to spend three
years of a PhD to
gather one data point of
information about one protein.
No, we can now
sequence our
entire genetic information, DNA
of a patient upon
diagnosis, after treatment, and
really understand how our medical
treatments affect
our biology on so
many levels. So we are living in
an era of biological
big data. But speaking to you, maybe
of course I have to say that in fact
there's only one true big data.
But here I really speak
about data
and biology that is
out there on all of these
scales that I've just
introduced to you. So
data is not a problem anymore,
but we need concepts how to
work with this data.
And in systems biology, there's two
sort of complementary
approaches to deal with all this
data. So again,
data can be on the level of organisms,
it can be on the level of cells, or it can
be on the level of molecules.
But the problem is you hardly
ever have just one of each of these.
So usually you have many
human beings in a clinical study, you have
many different cells that make up our body,
like 30 trillion of them
to be precise, or many
molecules that do something, right?
Many virus particles
or anything of that like.
So we have two different ways to approach them.
Either you do it sort of
bottom up, where you say, okay,
let's just look at
some of these different molecules
where we know actually what they are doing,
how they interact with one another.
And then we try to
sort of describe this
in a dynamic way,
so how they change over time. So first one
molecule interacts with another one, and then
these two interact with the third one.
And then we understand the small
part, like we start from the ground,
we understand the small part of it,
the small module, and then we sort of
subsequently try to
connect these modules.
And as I said, there,
based on a lot of knowledge, what is going on,
we have some mechanistic
understanding how these gear
wheels really fit into one another,
and then we can see how this system is changing
over time.
But then there's another way to
look at a biological system,
let's say a patient, a COVID-19 patient.
Just look at them from the
top, okay? You don't really know what's going on
in this patient, okay? It's a black box,
this patient. But now you give a certain
drug to this patient, this is
your input into the system, into your black
box, and you see how the patient reacts
to it. And this is your output, your reaction.
And then you do this not only with one, but with
the entire cohort of
COVID-19 patients, and you try
to see patterns, okay? Whenever we give
this to a certain patient, where maybe we have
all these health records,
there's always a good outcome.
Or maybe not a good outcome. And then you can start
to try to understand this input-
output relationship. Needless to say,
for this you need a lot of data, many patients,
many drugs that you
potentially test.
But then you can learn associations.
Okay, smoking
history is always connected
to a bad outcome, or as we said, a high BMI
is always connected to a bad outcome.
And these are all these machine learning, artificial
intelligence methods.
Okay, so these
are sort of the two complementaries,
not schools of thoughts, but approaches
that are not mutually exclusive, right?
Because you can, while looking at these
patients as a black box, you can still start
with some, for instance, cells of
the immune system that are attacked by the virus, you can also
start with some bottom-up modeling there, and then you
can again put it all together.
But to give you
a precise example,
I want to use the bottom-up
approach now as an example, okay?
So again, where we look at how
biological entities change over time.
And actually these models,
particularly also with these going ups and
down, these oscillations,
they were introduced a long time
ago for
for
foxes
and rabbits,
but biology is not just
that. And in fact, I wrote a
book
on
rabbit-free
biology. However, we can still
use all this mathematics that was established
to now bring
this research back to the future,
using this mathematics to understand something
about molecular biology.
And the last example
I want to give you,
before I hand over back to Noah,
is the example of a
molecule that controls
the survival of
our red blood cells.
It has this complicated
name, signal transducer and activator
of transcription, STAT5.
The only thing you need to know about this molecule
is it can bind
DNA, okay? It can bind our
genetic information and then
exert certain
tasks there. And if
we have it, our red blood cells are
surviving, and if this
molecule is not
at our genetic information doing its job,
then red blood cells
are dying. And so do we
at some point. So this molecule
basically shows us
a life-death decision,
which starts with a simple,
very simple question. Does 1
plus 1 always give 2? Because
if we have two of these molecules, they are supposed
to form this complex of two molecules.
And this sounds simple, but
actually biology is more complex than that, because
we not only have one
family member of this molecule, we have two
of them, the blue one and the red one.
And also this can be modified.
It can be chemically modified to be
switched. So this
small p here is a chemical group
that can be added to this molecule,
and then it's switched from the off state
by this addition of this chemical
to the on state.
And so we basically have four different
molecules that
potentially can do this job of finding
genetic information and making sure
our blood cells are surviving,
which gives us a matrix of potential
biological entities here.
Not this kind of matrix,
but more like this kind of matrix.
So from all these molecules, we can
form these potential complexes. And now
the problem is they are very small, so we cannot just
look at them to see which of them do actually exist
and what do they do.
But we do have
experimental data
where we see how these complexes change over
time. So what we see here
on the top is basically we see
how many of the switched on molecules
we have as compared to all
possible molecules. And then
we also see how the number of the blue
molecules compared to all molecules change
over time. And then here we basically
see how these two
ratios change with
respect to one another
in these cells.
We can measure this, and then with a mathematical
model, we can basically describe
this data.
As you can see, again, we have our
simulation and also the confidence
interval here. And now with
the simulation, we can basically look back and see
which kind of molecules
from this matrix do actually
exist. I mean, which one of those
we need to describe this data.
And the answer is surprisingly few
of them we only need.
And this is something we cannot just look at.
We really needed the
computer to find out what's the probability
that these molecules exist.
And this we also published. It's not
an A work, but again, A and B because
these are two molecules. And with this, maybe
we better understand now how our
red blood cells are surviving and
be happy.
To wrap this all up,
what I've shown to you is basically there are
standing parallels between computer viruses
and biological viruses.
We can apply computational algorithms
to address challenges that we are facing
in the COVID-19 pandemic.
And systems biology helps us to answer
complex biological questions with quantitative
data, for instance, how
molecular switches decide between life
and death in our blood.
And Noah will give you
some more teasers
how this all relates.
I'm just saying, if you
feel intrigued by this,
don't hesitate to reach out to us.
We will be happy to also
share more information.
Having said that, I hand over back to Noah.
Hi, and welcome back.
For anyone that
joined us
after my introduction, I will just
explain again that
my name is Noah.
I'm a Junior Master Scientist at the Weizmann
Institute of Science
here in Israel.
And my background
is from...
My background is
as most of you guys,
I was a reverse engineering and
cybersecurity expert at
Checkpoint for many years. And when I
joined Weizmann, I had to close
a huge
gap in my knowledge
in biology.
And I was...
What was stunning for me was the parallels
that I found so often in my
in the material that I
studied in the courses and also for the lab
with
the knowledge that I have from the past.
So, I will just give you
a few examples of
the things we have in biology
that will look very
familiar for you.
So,
without any further
introduction,
let's start with
cancer cells. Okay, so cancer
cells, unlike
most of the people think about them,
they are not just cancer
that accumulate in our body. This is
more like
cells in our
body that have superpowers. They evolve
better
than the rest of the cells in the body
due to a mutation
from a radiation,
from a UV light, or from
any other, from smoking,
or just
by chance in the body.
And then, what's happening with
these cells is that they become selfish.
Which means that instead of
working together
with the rest of the cells in the body,
they are selfish enough
to try to control
and to
get all of the resources, for example,
they are actually
sending signals to the
body to build
veins in order to get more oxygen
into the tumor environment.
And
one
stunning fact that I learned
during my courses
is that not only that the cancer
cells are selfish, they also
try to evade the immune
system, because
one of the roles of the immune system is
not only to
eliminate pathogens, but also
eliminate cells that are selfish, like
cancer cells. Because cancer cells, like we
have them on a regular basis in our
body, and
this is one of the main roles of the immune
system, to detect
them and
to eliminate them.
So, it's not
only that our immune system
is working as an antivirus, but
the thing here is that the cancer cells
are fighting back.
So, the picture here is representing
the interaction between cancer cells and different types
of immune cells. Cancer cells
are the one that shows in blue.
Immune cells are in different colors, and
what they are trying to do
is to
send... so, you can
think about the rest of the
colorful cells as
different types of antivirus from
many different companies,
or maybe
different models that we have in one
antivirus. And what the cancer cell
is trying to do is to actually reprogram
or to send
signals
to different immune cells
in order to evade
them. And this is something
that was, you know, when I heard about it,
it was really familiar for me. Like,
there were many models that tried
to evade my VM, to evade
my
AIDA, my
OLADPG, or the antivirus
in general. And
so, for me, it sounds pretty familiar.
So, apparently,
it wasn't our
idea to evade
people, to evade
entities that
are trying to detect us.
And I can promise you that cancer
cells are doing it in
a much more efficient way.
For example, one
very stunning
example that I have
is that
every cell is supposed to
present
a molecule
on its surface, which
the name of the molecule is an MHC.
And this molecule
is actually like
the info
that an executable
can have, which
describes what
these cells are producing,
what are the proteins that are running inside
of it. And cancer cells,
what they actually do is they just
don't show, they just don't present
these MHC molecules,
so it's really hard
for the immune cells
to understand what actually is going on
inside of them.
And
here
you can see something that
we call a hot tumor.
If a tumor is immunologically
hot, it means that the immune system
is operating here. However,
when cancer progresses, it learns
how to deactivate immune cells
and make tumors
immunologically cold. Like, look at this
tumor, and we can
see here many different
of
immune cells that are trying to detect
and trying to
get attached to it and to understand what's going
on inside. But after a
while, what we can see in tumors
that are actually
successful with evading the immune system
is that they become immunologically
cold tumors. You can see
that most of the
immune cells just gave up
and
they said, okay, whatever,
I have other stuff to do, and they just left
the tumor there alone.
This is why
you get
a tumor in your body.
And here is just
what you can see here
is just
a network of
different... this is just
a brief
knowledge that we have so far about
all the
different ways that cancer
has in order
to try to evade the immune system.
Generally, cells can communicate
with each other using diverse
signal transaction networks
that lead to different outcomes.
And what cancer cells do, they
upregulate the pathways which
inhibit normal
function of immune cells.
They don't even
just
tell the
specific immune cell, you know, just leave
me alone. They also ask him to spread
the rumor and tell
the rest of the immune cells, you know,
this guy, it's okay,
you can just leave him alone.
And this is something that was
really fascinating for me to learn.
And here
is an example
where cancers are
overexposing PD-L1,
a molecule that suppresses T-cell.
Important cell type involved
in cancer cell elimination.
On the right, you can also see
a reprogrammed immune cell called
tumor-associated macrophage,
which helps to spread
the inactivating signal instead of
fighting cancer. Okay, so
you can see that it's not only telling him, listen,
I'm okay, it also asks him,
it also asks these cells,
you can also tell the rest of the guys
that I'm fine and to leave me alone.
And this was amazing.
I don't know too many models that are
actually so
powerful and sophisticated
and it was amazing for me to learn.
And my next
example, which is actually one of my
favorites, is bacteriophage.
Okay, so
what is a bacteriophage? Here you
can see,
let's take a look
at this very beautiful
gif
of
like, okay, so
I'll just explain. First of all, bacteriophage
is like my favorite creature so far.
And what it's actually
doing, there is a creature
that we all know, it's known for bacteria.
But there is also another tiny
picture and
what bacteriophage is actually
trying to do is to
inject its DNA
into the bacteria
and then the bacteria, instead of
running their
own genome, it will run
the bacteriophage genome.
Does it sound familiar to anyone
else? Apparently also with this we didn't
come up. And this
was amazing for me to understand.
You can actually see it. This is actually
a bacteria,
and here you can see a bacteriophage
that are attaching
into it and injecting
their genome into the
bacteria. And here you
can see this is a very sad story because the
bacteria, sometimes they
understand that they are
doomed.
But this is just a gif.
So I will just explain a bit about
the progress of it.
So what you see here is that there is
attachment of the bacteriophage into
the bacteria. Bacteria is
the bacteria that
you know, that we all have in our body.
And this is the bacteriophage
which has a very, very
nice and interesting
picture and structure.
So what it's doing
is actually attaching into the bacteria.
It's injecting
its own genome
into the
bacteria's inside
surface. And then the bacteria,
instead of running, like you see, this is the
genome of the bacteria.
Instead of running its own code,
it's running the bacteriophage code.
And then the whole purpose of it
is that
there are more and more
bacteriophages that are
being formed inside of the bacteria.
And then
the poor bacteria, you know,
are just full of
bacteriophages until they release each other.
Until they get, like,
it's too much for it.
And they just go out
and infect even more and more
bacteria that are inside
of this environment.
And
another thing
that I'm pretty sure that most of you heard
about, if you are interested in
biology at all, is
something that is called CRISPR.
So we all know that CRISPR is being
used these days in order to
kind of engineering the babies
of the future and stuff like this. But what was
fascinating for me to learn is that
actually CRISPR, the whole
purpose of it, is
this is a defense
mechanism of the bacteria
in order to detect bacteria,
in order to detect phages.
So what you see here is
that this
pink evil thing is
the bacteria. And what
it's trying to detect, and
this is something that we actually
stole from them,
is what it's trying to detect is
the sequence that he also
that he
saw before
of phages that try to
attack it. And the way
that it's doing this
is just, you know, he has a sequence
that he remembers
and then he just tries to
do pairing, like to compare
the new
DNA that he sees
inside of the cell
and
the sequence that he already knows that
this belongs to the phages.
So, of course, I think
it's very
obvious that
this Cas9
system is very similar to what we know as
MD5, or the hashes of the
models that we have inside
many different
antiviruses.
And another
very cool thing is that it's not
only that the bacteria is detecting
phages, like
using the known
sequence, is that the phages
are actually invading this defensive
system by changing their genome.
The same way that malwares are changing
their MD5, you know, it's
enough to change just one bit.
Phages are doing the
same in order to
like, you know, via evolution
in order to being able
to inject
their code into new bacterias.
And
so this mechanism that we
actually stole from bacterias, we will
we are trying
these days to use it for genetic
engineering, but
just so you know,
it wasn't our idea, we stole it
from a very actually
smart creature, apparently.
And I will just
since our time is up,
I will just
say briefly what
the role of our specific
lab. So what you see is
so my research
is specifically about
the microbiome. What is the microbiome
is what we are trying
to do as you all
I don't know if you heard about it, but
we are all full of bacterias
inside of our gut and we
are trying in our lab
to understand
and maybe in the future
also to manipulate the different bacterias
that are inside of our gut and all the networks
that are inside of it. First of all
we try to find the connection between them
and different diseases like
diabetes and ALS
and obesity.
And so this is
a very new and interesting field
and you should know
that you all have bacterias
all over your hands, your
eyes and specifically inside your gut
and it's not just that they are there
happily not doing anything, they are
actually sometimes very crucial
in order to keep the
homeostasis.
So this was just
a brief thing
about the microbiome, just a brief
introduction to the microbiome.
Just a few things that we
found out so far in our lab is the
connection between the bacterias that are inside
the gut and the very well known
disease that are inside the ALS because
the bacterias, they can
send signals from the gut
into the brain and
there is a very
clear connection between
the dysbiosis that we will see in the
gut and
the form of this
disease. Here you
can see many different
types of diseases that are linked
to dysbiosis
in the microbiome.
And another very cool
subject we are
looking into in our lab is diabetes.
Just
one thing that I learned
and it was really interesting
and
like
this is one of my favorite challenges
in this field, is that
when you have any kind of disease, it's not like
that, for example diabetes, if you have
too much sugar in your blood,
the problem with it
wouldn't be
just about the fact that you have too much
sugar in your veins.
Many different diseases,
they are like, there is a hip
corruption, let's say,
in our body, but unfortunately
unlike in new operation
systems, in our body
there is no such thing
as different processes that
have different
classes of memory.
So if there is hip corruption
in one process,
in one system
in our body, it affects the rest
of it as well.
For example, in diabetes, one of the problems
that we see is that
we lose the
homeostasis of
also
the
CA2 inside our
cells.
And actually
we can see that
the blood is becoming acidic.
So this is just one thing.
so, I would just
say briefly that I'm really proud
really, really proud
that I'm taking part
of my lab.
The very nice person that you see
here is my supervisor. I have a professor
who is also a supervisor, and he
recently used me as
a lab rat. What you see
here is an insulin
sensor that
was attached to my shoulder for
around three weeks.
But I'm, of course, very proud.
I wish for
myself and for everyone always to be in the side
to help
patients get
a better cure.
And this is something that I'm really
happy about, that what I'm
doing here is actually helping
people that are sick.
But yes,
also,
all of us actually do stuff like this in the lab.
Also, the reason that
this guy is dressed like this is because
during the corona,
our whole lab became
of course,
we are not
doing anything that is linked
to
corona.
We didn't do anything
about it like before,
but during
this breakdown
of...