July 18, 2024 • 57 mins
In this episode of Inside Cancer Careers, we hear from Dr. Eytan Ruppin, Dr. Tiangen Chang, and Dr. Yingying Cao, computational scientists at NCI's Center for Cancer Research. They discuss their newly developed AI tool, LORIS, which can predict how a patient will respond to immunotherapy. LORIS analyzes six variables commonly measured in clinical settings to calculate a score indicating the patient's likelihood of response. They also share insights into the future of AI in cancer research and their career paths in science and offer advice to those interested in pursuing careers in biology, medicine, and computer science.
SHOW NOTES
Eytan Ruppin, M.D., Ph.D.
Tiangen Chang, Ph.D.
Yingying Cao, Ph.D.
Cancer Data Science Laboratory
NCI Center for Cancer Research (CCR)
Chang, TG., Cao, Y., Sfreddo, H.J. et al. LORIS robustly predicts patient outcomes with immune checkpoint blockade therapy using common clinical, pathologic and genomic features. Nat Cancer (2024)
LOgistic Regression-based Immunotherapy-response Score (LORIS)
Escher, Gödel, Bach: An Eternal Golden Braid by Douglas Hofstadter
AD
NanCI – Connecting Scientists
YOUR TURN RECOMMENDATIONS
Intelligence Squared (Podcast)
World Views, An Introduction to the History and Philosophy of Science by Richard DeWitt
New Scientist
Season 2: Episode 1 “The Future of AI and Cancer Research”
CENTER FOR CANCER TRAINING (CCT)
Inside Cancer Careers Podcast
CCT Website
LEARN MORE FROM THE NATIONAL CANCER INSTITUTE
Online
By Phone: 1-800-4-CANCER (1-800-422-6237)
U.S. Department of Health and Human Services
National Institutes of Health
National Cancer Institute
Mark as Played
Transcript
Episode Transcript
Available transcripts are automatically generated. Complete accuracy is not guaranteed.
(00:03):
[music]
Oliver Bogler (00:04):
Hello, and welcome to Inside Cancer Careers, a podcast from the National
Cancer Institute where we explore all the different ways people fight cancer
and hear their stories. I'm your host, Oliver Bogler from NCI's Center for Cancer Training.
Today, we're talking to three computational scientists working on cancer in NCI's Center
(00:24):
for Cancer Research. They've recently published an AI tool to help clinicians
predict response to immunotherapy, a great example of how people with all kinds of skills
can make important contributions to ending cancer as we know it.
If you recall, in our first episode of season two we talked to Tony Kerlavage and Jeff Shilling,
(00:44):
informatics leaders at NCI, about the general impact of AI on cancer research. Today we are
highlighting one of the teams at NCI working with AI, and exploring their patient-focused work.
Listen through to the end of the show to hear our guests make some
interesting recommendations and where we invite you to take your turn.
(01:04):
[music ends]
So with us today is Dr. Eytan Ruppin. He's the chief of the
Cancer Data Science Lab at NCI. Welcome, Eytan.
Eytan Ruppin (01:11):
Thank you. Thank you so much.
Oliver Bogler (01:13):
And also with us are two fellows in Dr. Rupin's group, Dr. Tiangen Chang. Welcome.
Tiangen Chang (01:18):
Thank you Oliver, glad to be here.
Oliver Bogler (01:20):
And Dr. Yingying Cao, welcome to you.
Yingying Cao (01:23):
Thank you, Oliver, with pleasure.
Oliver Bogler (01:26):
All right, Eytan, what is a cancer data scientist
and what role do they play in cancer research in 2024?
Eytan Ruppin (01:34):
So, as you may know, Oliver, now in cancer research and also in treatment,
we are generating a lot of data spanning different dimensions in a way that is exponentially
increasing every decade compared to the past. And that, like in many aspects of our life,
(01:57):
creates a large amount of data that really needs computers and computational analysis
to try to make sense of it, both to understand what's happening in cancer from a basic science
research point of view and to try and devise ways, which is the specific focus on my lab,
(02:19):
on trying to sift through the data and try to decipher what are the essentials
that can help us advance cancer therapy, both in terms of identifying new targets,
new combinations to treat patients, and in terms of predicting the patient's response to an array
(02:40):
of possible therapies, aiding the physician, of course not replacing the physician, but aiding the
physician to make clinical decisions and to choose the best matching treatment to any given patient.
Oliver Bogler (02:56):
So a great example of your work was recently published, Tiangen and Yingying,
you are the lead authors on that paper. It's focused on immunotherapy, a very fast moving and
important and promising area in cancer research right now. Please tell us about that work.
Tiangen Chang (03:11):
Sure, so we all know that cancer immunotherapy has been
a game changer in the cancer treatment over the past two decades. But here is the thing,
not every patient responds to this therapy. So on average, only 20 to 30 percent of the
patients will see a positive response. So there is an urgent clinical need to
(03:35):
really figure out which patients will eventually benefit from this therapy.
So LORIS is a machine learning model to help predict patient response to immunotherapy. And
it is a logistical regression-based model. It takes six variables of a patient as input and
(03:58):
these variables are routinely measured in current clinical settings. So output score
of LORIS can reflect the response probability of our patients to immune checkpoint blockade
therapies. So we have validated this model in a lot of currently existing cohort, but we
(04:18):
still want to see it to be further validated by more clinical cohorts in the future and finally
benefit more patients from this therapy and help clinicians to make better decisions.
Oliver Bogler (04:33):
So Yingying, how did you come to build the model in the way that
it is built? How did you choose the variables that Tiangen mentioned?
How did you put them together in a way that helps the clinician?
Yingying Cao (04:46):
The way how we choose the features is based on the how many
features are collected for each cohort. The most common features are the eight features
and we choose six of them to build our model. So because it is a model for clinical use,
(05:08):
we want to build a model that can be easily to use for the clinicians. So
it's more easier collected the features like the albumin, so it's collected from the blood
and the tumor mutational burden. So they are routinely collected during the clinical usage.
Oliver Bogler (05:30):
So these are readily available data items that most clinicians will have.
Yingying Cao (05:36):
Yes. Yes.
Oliver Bogler (05:36):
Explain to us what is logistical regression
or logistic regression. Did I say it wrong? I'm not a data scientist, have you noticed?
Tiangen Chang (05:44):
Yeah, it's correct. Yeah, absolutely correct. So yeah,
logistic regression is a kind of regression. So we know that linear regression, y equals ax plus b,
this is a linear regression. However, we know that the response rate is not from infinity,
(06:06):
right? It only have a range from zero to one. So you need to converge the from the
negative infinity to positive infinity into a range of 0 and 1 so here you need
to make the exponential and the division so here then you make the logistic regression
it is essentially a linear regression but just convert the value range from infinity to 0 to 1.
Oliver Bogler (06:31):
OK, I'm going to pretend that I understood that and ask a follow
up question and say, so how can I imagine it? I can imagine, and I've seen the tool. It's online,
right? Anybody can go to the web page and we'll put a link in the show notes. You can put in these
data elements. And then is the model weighting them differently and combining them into a score?
Tiangen Chang (06:52):
Exactly, so yeah you can input this I mentioned six features such as patient age,
the tumor mutation burden, the previous therapy history and so on. So you have the value for
these six variables then you just input these variables and there is an equation but although
(07:13):
you cannot see it but it's in the background then you click calculate this equation will take
these six variables as input and give you a value between zero and one. Say the output is 0 .5, that
means this patient have 50 % opportunity to respond to immune checkpoint blockade therapy.
Oliver Bogler (07:37):
Okay. Now you mentioned their mutational burden. I imagine
that as the sort of a measure of the number of mutations or the frequency
of mutations that you have observed in the cancer that comes from sequencing data.
Tiangen Chang (07:49):
Exactly.
Oliver Bogler (07:50):
Okay, and is that one of the more powerful variables in your model
or is it right there alongside some of these other features?
Tiangen Chang (07:58):
Yes, there's a little bit background. So this tumor mutation burden
is I think the first FDA approved biomarkers in cancer. So to quantify which patient can receive
this immune checkpoint blockade therapies. So FDA gives a threshold of 10. So that for patients,
(08:20):
if they have tumor mutation burden of 10 mutations per megabase on their genome,
they will be qualified to use this immune checkpoint blockade therapy.
However, later research found that this is a suboptimal biomarker because it will miss
a lot of real responders. That is there are a lot of, in our study we found that
(08:45):
maybe there are more than 60 % of the responders they have a tumor mutation
burden less than 10. So this biomarker is quite suboptimal and this is why we develop LORIS.
From LORIS we found that we can identify patients that are previously thought to be poor candidates
(09:06):
for immune checkpoint blockade therapies but they are actually the responders. So we hope
this tool can help clinicians to select more patients that can benefit from this therapy.
Oliver Bogler (09:20):
The model that you developed is based on historical data,
right? On retrospective data that you had. And I think you mentioned in the paper
that you need now as the next step or one of the next steps,
a prospective study to really test it. So tell us about that. What would that look like?
Tiangen Chang (09:39):
Yeah, so we have basically in clinic we have two types of data. One is
retrospective data, one is prospective data. Retrospective data that is collected before
that is already there. So you can actually you can do cheat, right? Because you already see the
result so you can modify your classifier so to fit the result so you can cheat. However,
(10:05):
for the prospective data you cannot do that because the data is still not there. So you set up
a new clinical trial and then you have the model first. Then you collect data to see if your model
really works or not. So prospective data is really important in computational science. So that can
evaluate if biomarker or findings really true or not because in this case nobody can do the cheat.
Oliver Bogler (10:32):
Interesting. So Yingying, a different question. I don't know,
when I looked at the paper, I read the paper and looked at the model,
it struck me as surprising that only six pieces of data can be so powerful
in that prediction. And that may be because I lack mathematical intuition, but were you
surprised by that? We always talk about cancer is so complicated, there's so many things happening.
Yingying Cao (10:55):
Yes, I think it's quite surprising because intuitively,
in the machine learning study, the more features should have better performance. But currently,
we found that these six features with the current data have the best performance.
I think it's not intuitive, it's surprising. The reason is that,
(11:21):
firstly, usually the clinical data is quite noisy. And when the data is quite noisy,
it doesn't mean that the more complex model or more features we have for the better results.
And another thing is that it doesn't necessarily mean because of the noise of the data and the
(11:45):
more features can cause overfitting. This is also a major, I think the major benefit of
our model that it can be generalized to test the data and the data set from other cohorts.
(12:07):
So I think it can be say surprising and it's not surprising because we can explain that.
Eytan Ruppin (12:14):
I would like to pitch in just briefly. So first of all, Oliver,
the simpler the model is, it's the better and it's more believable. So
simple is an advantage in the parlance of models, complexity, machine learning,
what have you. It's called Occam's razor. So simple is good. That's one.
(12:41):
Second of all, a lot of features are strongly correlated between them. So the fact that you
see only a few gives you a lot of information, hidden information also on other features.
The question is to pick the ones that are most informative on the
(13:03):
state of the system. And that is what very nicely Tiangen and Yingying have done.
But thirdly, we should remember that we think that if further tested and validated going forward,
this could be valuable. That's our hope, of course, our belief, our belief, okay? But we
(13:29):
are far from perfect. So the performance without going into technical details of measurements of
performance is 75%, 80 % accuracy, but not more. Okay? And we should remember that. And
that's a very important open challenge and we are not yet there. We are not there yet.
Oliver Bogler (13:54):
So it's a really good first step, but there's more to come.
Eytan Ruppin (13:58):
Exactly. And the way I would relate to it and even if prospectively verified and so on
is another way to help clinicians take the best decisions. It's not that you know the
futuristic whatever the AI medical robots that are now we are very far from that.
Oliver Bogler (14:21):
Right. So, Tiangen, the other thing I noticed is that this model has predictive powers
in a number of different cancers, right? It's not just for one type of cancer. And
I guess that points to a commonality of mechanism amongst these cancers in
relationship to the efficacy of the checkpoint inhibition. What are your thoughts on that?
Tiangen Chang (14:42):
Yeah, yeah, thank you for the great question. Yes. So yeah,
we, because we first of all, we collect a lot of cohorts, different cohorts, they have different
cancer types. So our first question is if we can build one model that can capture the commonality
between cancer types to get a model that predict the ICB response in different kinds of types. So
(15:08):
then that is our first model. And that as you may know, this model captures the commonality
but does not capture the specific special property of individual cancer types. So then we developed
the second LORIS model that is specifically lung cancer. We’re doing lung cancer because this is
(15:29):
a cancer type that we have the largest sample. So we want to still keep the statistic power.
Then we found that using the same strategy we can also predict the lung cancer immunotherapy
response quite well. But the model is slightly different and the lung
cancer we added one feature that is PDL1 expression. And this is a commonly used
(15:53):
feature in lung cancer. Clinicians already know that the PDL1 expression is a very
important feature that determines the immunotherapy response.
Oliver Bogler (16:05):
Right, PD -L1 is one of the checkpoints, right? The checkpoint
inhibitors try to knock down, so to speak, to allow the immune system to fire. Is that right?
Tiangen Chang (16:15):
Exactly, yeah, it is expressed on the cancer cells and stop T-cells to kill them.
Oliver Bogler (16:20):
Eytan, thinking about the LORIS work and the many other things your group does,
and you just told us we're not about to have medical robots coming into our hospitals,
what do the next years hold for your AI work and AI in general in this domain? What's coming up?
Eytan Ruppin (16:41):
Yeah, that's a fantastic question and I do not know the answer. I can tell you what
we are working on and excites me and I'm hopeful about. So we is my lab and our collaborators of
(17:02):
course. So we are working on two I am gung ho about developing methods to better match
treatments to patients. Okay? And I'm trying to do that in two main new dimensions. One of them is,
(17:24):
believe it or not, look at the good old hundred years pathology slides, you know, which are the
backbone of tumor classification and and via that some treatment recommendations and actually apply
the recent developments, buzzwords in AI and machine learning and deep learning and all that,
(17:49):
and predict the molecular characteristics of tumors directly from the pathology slides. The
gene expression, the proteomics, genomic alterations, mutations, what have you.
This is now an extremely hot field with a few fantastic labs working in the US,
(18:11):
Europe, whatever. And we are also heavily excited and invested in that. Actually,
we just published a paper in Nature Cancer today on this topic. And a
month ago in Nature Medicine, so you can see the pace, the pace of things.
(18:32):
And in this paper that we published today, we show that we can actually take the slides and bring,
predict omics data from them. And based on that, in previous work that we have done,
bring three recommendations. And we are showing validations in five different
independent cancer types and treatments. So that's one direction. Slip sliding again,
(18:58):
Oliver, if you remember Paul Simon's, I mean, “the kids” have no idea what we are talking about,
but right, the young researchers. So that's one direction, which I'm gung ho. What can
we do from the slides? And many others, of course, others doing fantastic work on it.
(19:21):
The other topic which Tiangen and Yingying are actually taking active part in
is asking what can we learn from the blood on the tumor immune microenvironment? And so
(19:41):
together with other wonderful collaborators in the NCI, we are collecting matched data
where we have from the same patients, both the blood and tumor molecular information.
We are doing what is called single cell sequencing and what have you and whatever.
And we are learning what we can predict from the blood about what's happening in
(20:07):
the tumor microenvironment. And for that, for example, predict the response to immune
checkpoint therapy. So this guy, Tiangen here, has now a paper finding with some
very exciting findings on that which we are going to submit in the next few days.
(20:29):
And you know, if I'll tell you, I'll have to kill you, dear.
Oliver Bogler (20:34):
I won't ask.
Eytan Ruppin (20:36):
So anyway, so that's the two main directions where, you know,
I'm excited about and I believe that we can deliver meaningful advancements in the next
few years. And here, this is recorded so you can call me accountable, right?
Oliver Bogler (21:00):
Yeah, we'll check in in a few years and see how you did.
Eytan Ruppin (21:03):
The Virtual Site Visit.
Oliver Bogler (21:05):
Exactly. Those are exciting areas. Sounds fantastic. Tiangen, I know I can't ask you
specifics about the paper you're about to submit. It sounds really interesting. But
is it building on LORIS? Is it a natural? No, it's something completely different. Tell us.
Tiangen Chang (21:23):
Okay, yeah. So yeah, currently I actually, yeah, I continue working on predicting
and understanding the response to immunotherapy. But now I'm more focusing on studying the tumor
microenvironment. So I do not limit myself to predict the immune checkpoint blockade response
(21:44):
only from clinical data. So for the paper Eytan just mentioned, so yeah, we use more advanced
techniques such as single-cell RNA-seq so we can look at individual immune cells and tumor
cells in the tumor microenvironment and also in the blood. So we use a similar strategy. As
(22:06):
Eytan said, we computational scientists, we collect data from public databases.
So we collect a huge cohorts of data. To our knowledge, it is the largest head and neck
cancer. This is a single cancer type. And we found the B-cell signature predict the immune checkpoint
(22:28):
blockade response quite well. And it only, not only works in the tumor microenvironment,
but also works in the blood. So simply from get the blood of our patients,
then do a single cell or even simpler, do the flow cytometry. We can predict
the response probability of this patient to immune checkpoint blockade therapies.
Oliver Bogler (22:51):
Right, so this is a kind of a new phase where you're not using clinical data,
you're using more biological data that are perhaps also harder to acquire than routinely
perhaps collected in the clinic. But is there a future where the two might converge and allow you
to have a more comprehensive prediction model? I mean, Eytan mentioned that the LORIS model
has an accuracy of approximately 80%. So will you get to higher numbers with this approach?
Tiangen Chang (23:17):
Yes, we hope so. Yes, one of the advantages of machine learning and this
artificial intelligence is to integrate data from different modalities, from
the clinical data, from the molecular data, and to integrate all this data. So we hope
we can increase the predictive power to lead us to a more accurate prediction.
Eytan Ruppin (23:43):
So, yeah, so I want to mention in that aspect, you know, it's very hard,
Oliver, there is little molecular analysis of patients' data yet because of the costs and
because it's not part of routine practice. So, where we hope that it could be a game changer is
(24:04):
these technologies where we look at the slides and infer the molecular data and we can do it
pretty well now. That can potentially enable us to contribute to democratizing medicine.
So you know, even within our nation, there's huge differences between different regions and
(24:26):
so on and so on. It is our hope that if further developed, this could be a very important factor
and I truly believe we can do it. I am actually, together with Ken Aldape, our chief of pathology
and my good friend and collaborator, we connected a few foundations that do work in Africa. And
(24:50):
we are actually engaged almost every week. We spend time meeting with at least one or two groups
enlisting them to get data from Africa, deploy our methods. There's a whole host of problems and
challenges. It's a topic for another discussion and interview. But we are persistent people.
(25:16):
So actually, I grew up as a kid in East Africa. And I actually don't test my Swahili,
but I actually have a deep connection and a moral commitment is a strong word,
but whatever interests. So I'm very excited to see how it will unfold. And the people we speak to,
(25:43):
the medical oncologists and the pathologists in Africa are really excited about that. I
hope we can look forward and make a difference. It's early days. Yeah.
Oliver Bogler (25:53):
But that's exciting. So the humble H&E slide may put the
sequencing companies out of business.
Eytan Ruppin (25:59):
Never. That's my plan. That's my plan. You know,
the revenge of a lowly federal clerk.
Oliver Bogler (26:10):
But exciting, as you say, it will make that kind of information available
in many more parts of the world where the kind of deep sequencing is still out of reach. That's
exciting. And then I understand those molecular data inferred from the H&E slide could then be
combined into predictive models that you're building. Yingying, how does your next piece
(26:32):
of work, your next project fit into this scheme now that LORIS has been published?
Yingying Cao (26:37):
As you know, that LORIS used the features from the clinical routine collected data,
and they cannot explain more details in the mechanism of how the immunotherapy
response in each patient is and how the tumor in the human body,
(27:00):
the tumor microenvironment react to the tumor cells. So I'm more interested in the details of
the interaction of tumor microenvironment and the tumor cells. So currently I use the single-cell
RNA-eq data and the spatial transcription data to really understand the tumor microenvironment.
Oliver Bogler (27:27):
Fantastic. I'm so excited to hear about all the things you're all
doing and I can't wait to see what amazing advances you make. We're going to take a
quick break and when we come back, we will be talking about career paths with our guests.
[music]
Cancer researchers, are you drowning in a sea of scientific papers? Wish there was a way to
(27:49):
cut through the noise and find the research that truly matters to you? Look no further than NanCI,
the revolutionary new AI-powered app from the National Cancer Institute.
NanCI isn't just another research publication app. It's your personal research assistant,
designed to understand your unique interests and connect you with the most relevant papers, people,
and events in the field. Imagine being able to chat directly with research papers, ask questions,
(28:15):
and get instant answers. That's the power of NanCI's AI-driven "chat with paper" feature.
But NanCI is more than just a conversation starter. It's a networking tool, helping
you discover new connections and build your professional network. It's a learning tool,
recommending papers you might not have found on your own and helping
you quickly get up to speed in new research areas. Simply declare an
(28:35):
interest by bookmarking papers on a topic into a folder and NanCI helps you stay up to date.
Whether you're a seasoned researcher or early in your career, NanCI can
help you take your work to the next level. By harnessing the power of AI,
NanCI is transforming the way cancer research is done.
Download NanCI for iOS today from the Apple App
Store! That is N-a-n-C-I. Experience the future of research for yourself.
(29:02):
[music]Oliver Bogler: All right, we're back. Eytan,
you did an MD and a PhD in computer science, an unusual combination. How did that happen?
Eytan Ruppin (29:12):
Serendipitously… or whatever you say it in English, right? So actually my
favorite talk to young fellows in the NCI is the title is ‘how not to build a scientific career’.
All the mistakes I made and there are so many Oliver, there you'll be. So yeah, I started late,
(29:39):
I studied medicine. And then at the age of 30, I finished my studies, finished my internship
and I came to the conclusion that I do not want to be a practicing physician. And then
what will I do, right? And so I was really thinking, I was quite lost, to be honest.
(30:08):
And then I came upon a book, because you asked us to think about the book
or movie or something that influenced our lives. And that is a book that changed
my life. So the book is called Bach, Escher, Gödel by Douglas Hofstadter.
Oliver Bogler (30:27):
Yes, I've read it, yeah.
Eytan Ruppin (30:29):
Yeah, okay. Did it change your life?
Oliver Bogler (30:30):
It didn't change my life. Yeah, but long time ago.
Eytan Ruppin (30:32):
So that's an amazing book about the deep connections between computer science
to music and art and the concept of recursion, but brilliantly written. And I decided that I
want to become a computer scientist. So at my time, you know, all the dinosaurs
(30:52):
roamed the earth and you remember them, they were big and green and frightening,
right? There wasn't an MD, PhD or something in my country. So I had to do a first, second and
third degree in computer science in parallel to my residency in psychiatry. So I didn't sleep
(31:17):
for about five or six years. And luckily I worked hard and did it in record time. And
then I went to a post-doc and a residency in the US and the rest is history.
Oliver Bogler (31:35):
Interesting. Yeah, you founded the Cancer Data Science Lab at NCI, right?
Eytan Ruppin (31:40):
Correct.
Oliver Bogler (31:41):
And we'll get back to that in a moment. But first I want to ask Yingying,
what got you into science in the first place? What was that moment
that made you say, I want to do computer science?
Yingying Cao (31:53):
During my bachelor study, I studied preventive medicine and
I even don't know what bioinformatics is after I graduate from bachelor degree. I was enrolled
into a bioinformatics group during my master degree study. And I found it fascinating that
(32:15):
because in preventive medicine study, we focus on the population medicine. We focus on how to
make some strategies to prevent the disease. For example, the change of the living style,
(32:35):
just to make some strategies to do on the population level. So we don't do it on the
individual level. And during my bioinformatics study in my master degree, I found that each
human body can be quantitatively, can be a quantitative object. So each molecules in the
(32:59):
human body and even the abundance of different cell types can represent your immune level.
So as I found it's quite interesting to treat each human body as a quantitative object and
to see how the human body can fight off the infectious disease. The math degree I studied
(33:24):
the virus infection, infectious disease. Actually the cancer research and the infectious disease
they share very common things, especially in the immune immunity. So I found that immunity
is a very fascinating subject. And then I entered NCI to study the cancer research,
(33:50):
the cancer immunotherapy. And I found it, yeah, I think every step is unexpected. When I studied,
I found it interesting. And so I don't know what the next step is, but I found I'm more and more
interested in the scientific research. And I want to build my career on the scientific research
(34:18):
to be in the future maybe a mentor because during my scientific past, I meet a lot of
very good mentors. I want to transfer what they teach me to the trainees. And also,
I also have a lot of ideas even now that, but I cannot do it alone. So I want to build a
(34:44):
scientific team to include different talents and to collaborate and to...
find the truth about immunity, the cancer research, and even some infectious disease.
Oliver Bogler (35:00):
Thank you for sharing that. So you're kind
of immunity first and then you pick up the tools you need as you go along. Tiangen,
how about you? What got you interested in science and computer science?
Tiangen Chang (35:13):
Sure I think yes it's a great question and the answer for me is
simple I think it's largely due to personality so when I was young I can spend a lot of time
sit down to work on a problem for a very long time and I was not very good at playing games
(35:33):
or socialization with people and I enjoyed reading walking in the nature or exploring
science. So I don't like doing repetitive work but I found a great joy in spending a lot of
time thinking about a problem and eventually finding the answer. So that moment of discovery
(35:58):
makes me so happy and I enjoy that feeling immensely.
So actually I started my PhD career in plant science before that I have a master bachelor
degree and at that time I knew little about biology. So I didn't find much
difference between animal science and plant science. My PhD supervisor was wonderful.
(36:22):
He taught me two very important lessons that I still benefit from it. So firstly,
you need to trust yourself and focus on important questions. Do not just follow others. Secondly,
focus on translationally focused research. I'm not saying that basic research is not good,
actually basic research is extremely important but you know every supervisor
(36:47):
has their own style and this style will influence greatly on their trainees.
At 2021 I decided to study cancer this year one of my uncles diagnosed with cancer and before
that two of my grandfathers also had died of this disease. So I decided to devote myself
(37:13):
in fighting cancer. And I always feel very fortunate to have the opportunity to get into
this field. So I feel very thankful for Eytan. And even though I don't know why he chose me,
because at that time I was a completely clueless guy. Maybe that day he was drunk,
(37:34):
but that was a really beautiful mistake. I also want to thank Yingying, she joined
Eytan’s lab a little bit earlier than me and he introduced Eytan to me. So this is my story.
Oliver Bogler (37:48):
Thank you. Thank you for sharing that. I mean, this is really interesting to me.
So all three of you really started in biology and in medicine in different parts of that ecosystem,
but came to data science more out of an interest or a recognition that that tool set and that way
of approaching a problem could be interesting and helpful. Right? Is that, did I get that right?
Eytan Ruppin (38:13):
Yes, but not exactly in my case. So I...
Oliver Bogler (38:15):
Okay. I mean, you got lost and then you found yourself.
Eytan Ruppin (38:19):
I'm still drunk according to Tiangen.
Oliver Bogler (38:21):
Ha ha ha.
Eytan Ruppin (38:22):
So no, so actually, you know, my career has been so tumultuous. So I actually,
because I did my residency in psychiatry, I was naturally interested in understanding
what's happening in the brain, you know, in neurological and psychiatric disorders.
(38:42):
So I actually spent my first 15 years of my career as a computational neuroscientist building models,
neural network models, understanding, trying to understand what happens in Alzheimer's,
schizophrenia in the brain. And after, I don't know, I was already a professor of
medicine and computer science accomplished blah, I came to realize that I will not
(39:11):
understand it. I don't believe that in my lifetime, anyone else will understand it.
So I decided to take the bold step and reinvent
myself completely and move to a much easier problem, how to solve cancer.
Oliver Bogler (39:27):
Fair enough.
Eytan Ruppin (39:28):
That's true, that's true.
Oliver Bogler (39:29):
Yeah? And how's it going?
Eytan Ruppin (39:31):
Not so bad, I mean, surprisingly good. I mean, you know, our ability to understand
what's going in cancer and to come up with better treatments is advancing, I think, objectively or
not objectively at a much more impressive pace than what we understand about the brain how
(39:56):
the brain works. So I'm very, very happy that I finally made one good choice in my career. Yeah.
Oliver Bogler (40:09):
I'm sure you made many, but we're very excited that you chose cancer
because we need incredible people like you and your team. So I wonder as my final question,
what is your advice to people who are listening, who might be primarily interested in biology,
medicine, cancer, and be thinking about, well, how do I continue to do that in the current AI
(40:31):
revolution? Or conversely, coming from the computer science side and thinking about how
do I come into the world of cancer? I want to do something really important and meaningful
in my life. What's your advice to people, young people thinking about these questions?
Eytan Ruppin (40:47):
That's a wonderful question. I'll be happy to give you my take,
but I want first to give a chance to my fellows to answer it, and then I can take it from there.
Oliver Bogler (40:57):
Please.
Tiangen Chang (40:59):
Yeah, I think yeah from my experience you need to first to choose
a important question, so research question, then you just focus on this question. I still
remember that three years ago when I planned to join Eytan’s lab I prepared a research
plan so I just want to appeal to attract Eytan’s to see how beautiful this plan is,
(41:25):
how thoughtful I was and so he chose me. But yeah, now when I look back,
it's very naive. As Eytan said, all is very good. Only problem is that it cannot be realized.
But I found that still the question is still there. So at that time,
I just want to work on the cancer immunotherapy and the tumor microenvironment. And I'm still
(41:51):
working on it. And I will work on it for the next five, 10 years.
So I think the scientific question is really important, but the method will change a lot
because with the advancement of measurement and the computational tools. So for me, I never put
(42:12):
a major of my attention on the technical things. I put more attention on the scientific questions.
Oliver Bogler (42:18):
Yingying, what's your perspective?
Yingying Cao (42:20):
Yeah, I agree with Tiangen that I think the scientific question
in your mind that the puzzle you, that motivates you is very important. If you
don't have that question in your mind that you want to solve it, it's difficult to continue
because you will meet a lot of challenges and difficulties during the scientific research. If
(42:46):
there is no scientific question that h as you and the kind of mud can make you sleepless,
it's difficult to continue. So I think it's very important to find a real interesting question.
For me, it is really fascinating the tumor micro environment and, I'm very lucky. I think this area
(43:13):
is a very lucky area for bioinformaticians because more and more advanced techniques
and currently the spatial techniques can give you the information, the spatial information
of the molecules’ and the cells’ locations and even the distance between different cells. And
(43:36):
it's unbelievable maybe 10 years ago and even people are interested in it. It is difficult to
study that in such details. And also we can do it in single cell level and the spatial level. And
so I think more and more advanced techniques can help us to understand our questions in our mind.
(44:01):
And there is one part I don't agree with Tiangen is that Tiangen don't think the
techniques develop algorithms or techniques are important. I think the algorithms and the
techniques are very important. And the techniques are very important,
(44:21):
but the algorithms are even more important because we need to understand more reliable to interpret
the data. If the tool or the algorithms we rely on is not reliable and the interpretation of the
data in the prediction of the biological system is wrong. So I think that if I develop a team in
(44:47):
the future, I will focus on the algorithm development and also the biological part.
Oliver Bogler (44:55):
Thank you so much. Thank you. Eytan,
what do you say to people thinking about a career in your area?
Eytan Ruppin (45:03):
Yeah, so I will answer at the higher level. Okay. You know,
at my rank, we are clueless. We only talk at the higher level, right?
Oliver Bogler (45:15):
Of course.
Eytan Ruppin (45:16):
So seriously, you know, only choose this career if you love it. If you're excited by
science, if you're excited about the cause, you know, listen to your heart. Like in any decision
we make in life. You can think from here to the moon and back but listen to your heart.
(45:40):
Second of all, be a nice person. It is so important for your success. Science
today is a team work, both in running your own lab
and creating a good atmosphere. I always say I only hire and work with nice people,
(46:05):
okay, because I'm not paid enough as a lowly federal clerk to work with not nice people. No.
And also, with the collaborators. It's so much emotional IQ, how it's called, you know,
in leading, not less important than every other aspect. So be a nice person, an honest person,
(46:30):
a good collaborator. You know, we are at the data science lab when I was hired by
Tom Misteli. You know, he asked me, what's your vision, Eytan? And I said, you know, we will be
a center. We will help people. And I believe we do. We are not only doing our own research,
(46:50):
of course we are, but we are helping a lot of other people and other people are helping us.
So that's the second thing. First one was love. The second one was goodness and kindness. And
the third one is bravery. Be brave, okay. Pursue important questions. Don't strive,
(47:19):
especially it's a tendency, I don't know, in young PIs or whatever, to publish a lot. That's stupid.
Try to do something meaningful. And that is especially true in CCR at the NCI,
okay? Our mandate, explicit mandate, as you know, Oliver, is to pursue high risk, as much as we can
(47:43):
projects and be persistent and relentless and try to solve them, you know, on a long range.
So just to close circle, I told you that today we published a paper and so on. This paper actually,
without going into the details unless you ask me, closes 12 years of work,
(48:07):
of a continuous work in my lab on a specific challenge. We started in
2012. Our first paper was published in Cell, so not in a trivial zone,
in 2014. Since then, we have been working tirelessly to try and improve that.
(48:28):
Okay, so the starting point was published in Cell, okay, but wasn't good enough. So luckily,
I always say to my fellows, we are so bad that we can always improve. Okay?
So because the challenges are so big, okay, so 12 years of research I was just telling
(48:50):
my wife this morning, you know, I will not mention her name because she will kill me,
that you know we published this paper and it's 12 years of the fruits of 12 years of
work. So that, you know, that makes me feel take a pause and think about it.
Oliver Bogler (49:13):
Yeah, yeah, yeah, fantastic. Well,
thank you all for that advice. Really, really good and thoughtful.
[music]
So it's time for our segment now that we call your turn,
right? Because it's a chance for our listeners to send in a recommendation that they would like to
(49:33):
share. If you're listening, then you're invited to take your turn. Send us a tip for a book,
a video, or a podcast, or something that you found interesting or inspirational or amusing.
You can send those to us at NCIICC@nih.gov. Record a voice memo,
send it along, and we may just play it in an upcoming episode. But I'd
like to invite our guests to take their turn. Let's start with you, Yingying.
Yingying Cao (49:55):
I will introduce a very interesting podcast. Its name is Intelligence Squared US. I
don't know whether you're familiar with it. I found it very interesting. It's about the
debate on the important topics, including some science and technology and the society
and the culture. And I think the different topics, as you can imagine in your life.
(50:23):
I'm very interested in listening to the debates. I found for one topic when they raise that topic,
I have some opinions in my mind. But after I listen to their debates from different part,
I change my mind sometimes and the topic become more clear to me.
(50:49):
I think this is very familiar with the scientific research. Yeah, this is what
scientific research make me interested because I always want to solve puzzles and make everything
in my life more clear. So for example, if I meet something or some challenges in my life,
(51:16):
I want to make it I want to understand it and make it clear. And the Intelligent Squared
US is a debate that can make things clear. And it make you think twice
and always step back and think twice, step back and think twice and make this clear.
And I think it's a very interesting podcast and people can learn a lot.
Oliver Bogler (51:39):
Thank you. Thank you for that tip. We'll put a link
in the show notes. Tiangen, what have you got for our listeners?
Tiangen Chang (51:44):
I would like to recommend a book and the title of the book is World Views,
an introduction to the History and Philosophy of Science. The author of the book is Richard
DeWitt and this book is a great read that takes you on a journey through
the evolution of our understanding of the world from ancient times to the modern day.
(52:06):
This book really blew my mind by showing how different world views can lead to
completely different understanding of the same universe. And after reading this book,
I have had a feeling that even now most of our scientific findings could still be wrong. But
they are meaningful because they represent the steps towards a better understanding of our
(52:29):
world in the future. Just like a famous saying that all models are wrong, but some are useful.
So finally, I want to highlight that what is really cool about this book. So this
book just breaks down very complex ideas in a way that is extremely easy to follow.
So it is perfect for anyone who is curious about our world, science and its history.
Oliver Bogler (52:54):
Thank you. That sounds really great. I might just pick
up a copy. That sounds really interesting. Eytan, your turn.
Eytan Ruppin (53:02):
Yeah, so I just work guys. I don't read anything. I don't listen. I have
no time. I don't understand. I'm a sucker of popular science and science fiction. So,
you know, very large genre and I don't know if you know, Isaac Asimov, who is,
(53:24):
and so on, actually, in addition to being a fantastic science fiction writer and,
you know, the series of the robots and, you know, other series, he also wrote fantastic
popular science books. Really fantastic. Once when I had time, I would read them.
(53:51):
What I do now is I subscribe to New Scientists. New Scientist is in my mind a fantastic popular
science. I read a lot. I'm an aficionado of physics and in another world I would like to
be a physicist. Somehow it didn't, wasn't in the cards for me in this life, but maybe in another
(54:16):
life because I'm so fascinated. And our youngest son studies physics, so I live vicariously by him.
And my favorite thing is to watch different physics, popular physics, you know,
shows in series with him whenever he comes home. And then he explains to his stupid dad, you know,
(54:41):
very complex concepts. And he's very patient with me. And I'm so happy. I feel so proud.
I'm so happy to be stupid with him. You see what I mean? And I live vicariously. So I recommend
(55:01):
New Scientist. I'm not a show shareholder. I have no conflict of interest. I love popular science.
Oliver Bogler (55:13):
I think that's a really good recommendation. Thank you very much. Well,
thank you all three of you for spending time with us and coming on the pod and
sharing your insights and about your exciting work. I wish you all nothing
but the most tremendous success. You're really doing such important work. And so thank you.
Eytan Ruppin (55:32):
I want to thank you really very much for having us. And I'm not just
being polite. You know, I come from the Middle East, I'm not polite. By definition,
right? But I always talk sincerely. So I think it was very enjoyable,
I hope my colleagues, Yingying and Tiangen, shared the same thing.
(55:56):
[music]
Oliver Bogler (55:56):
That’s all we have time for on today’s episode of Inside Cancer
Careers! Thank you for joining us and thank you to our guests.
We want to hear from you – your stories, your ideas and your feedback are welcome.
And you are invited to take your turn and make a recommendation
to share with our listeners. You can reach us at NCIICC@nih.gov.
(56:17):
Inside Cancer Careers is a collaboration between NCI’s Office of Communications and Public Liaison
and the Center for Cancer Training. It is produced by Angela Jones and Astrid Masfar.
Join us every first and third Thursday of the month wherever
you listen – subscribe so you won’t miss an episode.
If you have questions about cancer or comments about this podcast,
(56:41):
you can email us at NCIinfo@nih.gov or call us at 800-422-6237. And
please be sure to mention Inside Cancer Careers in your query.
We are a production of the U.S. Department of Health and Human Services,
National Institutes of Health, National Cancer Institute. Thanks for listening.
[music]
Oliver Bogler (00:04):
Hello, and welcome to Inside Cancer Careers, a podcast from the National
Cancer Institute where we explore all the different ways people fight cancer
and hear their stories. I'm your host, Oliver Bogler from NCI's Center for Cancer Training.
Today, we're talking to three computational scientists working on cancer in NCI's Center
(00:24):
for Cancer Research. They've recently published an AI tool to help clinicians
predict response to immunotherapy, a great example of how people with all kinds of skills
can make important contributions to ending cancer as we know it.
If you recall, in our first episode of season two we talked to Tony Kerlavage and Jeff Shilling,
(00:44):
informatics leaders at NCI, about the general impact of AI on cancer research. Today we are
highlighting one of the teams at NCI working with AI, and exploring their patient-focused work.
Listen through to the end of the show to hear our guests make some
interesting recommendations and where we invite you to take your turn.
(01:04):
[music ends]
So with us today is Dr. Eytan Ruppin. He's the chief of the
Cancer Data Science Lab at NCI. Welcome, Eytan.
Eytan Ruppin (01:11):
Thank you. Thank you so much.
Oliver Bogler (01:13):
And also with us are two fellows in Dr. Rupin's group, Dr. Tiangen Chang. Welcome.
Tiangen Chang (01:18):
Thank you Oliver, glad to be here.
Oliver Bogler (01:20):
And Dr. Yingying Cao, welcome to you.
Yingying Cao (01:23):
Thank you, Oliver, with pleasure.
Oliver Bogler (01:26):
All right, Eytan, what is a cancer data scientist
and what role do they play in cancer research in 2024?
Eytan Ruppin (01:34):
So, as you may know, Oliver, now in cancer research and also in treatment,
we are generating a lot of data spanning different dimensions in a way that is exponentially
increasing every decade compared to the past. And that, like in many aspects of our life,
(01:57):
creates a large amount of data that really needs computers and computational analysis
to try to make sense of it, both to understand what's happening in cancer from a basic science
research point of view and to try and devise ways, which is the specific focus on my lab,
(02:19):
on trying to sift through the data and try to decipher what are the essentials
that can help us advance cancer therapy, both in terms of identifying new targets,
new combinations to treat patients, and in terms of predicting the patient's response to an array
(02:40):
of possible therapies, aiding the physician, of course not replacing the physician, but aiding the
physician to make clinical decisions and to choose the best matching treatment to any given patient.
Oliver Bogler (02:56):
So a great example of your work was recently published, Tiangen and Yingying,
you are the lead authors on that paper. It's focused on immunotherapy, a very fast moving and
important and promising area in cancer research right now. Please tell us about that work.
Tiangen Chang (03:11):
Sure, so we all know that cancer immunotherapy has been
a game changer in the cancer treatment over the past two decades. But here is the thing,
not every patient responds to this therapy. So on average, only 20 to 30 percent of the
patients will see a positive response. So there is an urgent clinical need to
(03:35):
really figure out which patients will eventually benefit from this therapy.
So LORIS is a machine learning model to help predict patient response to immunotherapy. And
it is a logistical regression-based model. It takes six variables of a patient as input and
(03:58):
these variables are routinely measured in current clinical settings. So output score
of LORIS can reflect the response probability of our patients to immune checkpoint blockade
therapies. So we have validated this model in a lot of currently existing cohort, but we
(04:18):
still want to see it to be further validated by more clinical cohorts in the future and finally
benefit more patients from this therapy and help clinicians to make better decisions.
Oliver Bogler (04:33):
So Yingying, how did you come to build the model in the way that
it is built? How did you choose the variables that Tiangen mentioned?
How did you put them together in a way that helps the clinician?
Yingying Cao (04:46):
The way how we choose the features is based on the how many
features are collected for each cohort. The most common features are the eight features
and we choose six of them to build our model. So because it is a model for clinical use,
(05:08):
we want to build a model that can be easily to use for the clinicians. So
it's more easier collected the features like the albumin, so it's collected from the blood
and the tumor mutational burden. So they are routinely collected during the clinical usage.
Oliver Bogler (05:30):
So these are readily available data items that most clinicians will have.
Yingying Cao (05:36):
Yes. Yes.
Oliver Bogler (05:36):
Explain to us what is logistical regression
or logistic regression. Did I say it wrong? I'm not a data scientist, have you noticed?
Tiangen Chang (05:44):
Yeah, it's correct. Yeah, absolutely correct. So yeah,
logistic regression is a kind of regression. So we know that linear regression, y equals ax plus b,
this is a linear regression. However, we know that the response rate is not from infinity,
(06:06):
right? It only have a range from zero to one. So you need to converge the from the
negative infinity to positive infinity into a range of 0 and 1 so here you need
to make the exponential and the division so here then you make the logistic regression
it is essentially a linear regression but just convert the value range from infinity to 0 to 1.
Oliver Bogler (06:31):
OK, I'm going to pretend that I understood that and ask a follow
up question and say, so how can I imagine it? I can imagine, and I've seen the tool. It's online,
right? Anybody can go to the web page and we'll put a link in the show notes. You can put in these
data elements. And then is the model weighting them differently and combining them into a score?
Tiangen Chang (06:52):
Exactly, so yeah you can input this I mentioned six features such as patient age,
the tumor mutation burden, the previous therapy history and so on. So you have the value for
these six variables then you just input these variables and there is an equation but although
(07:13):
you cannot see it but it's in the background then you click calculate this equation will take
these six variables as input and give you a value between zero and one. Say the output is 0 .5, that
means this patient have 50 % opportunity to respond to immune checkpoint blockade therapy.
Oliver Bogler (07:37):
Okay. Now you mentioned their mutational burden. I imagine
that as the sort of a measure of the number of mutations or the frequency
of mutations that you have observed in the cancer that comes from sequencing data.
Tiangen Chang (07:49):
Exactly.
Oliver Bogler (07:50):
Okay, and is that one of the more powerful variables in your model
or is it right there alongside some of these other features?
Tiangen Chang (07:58):
Yes, there's a little bit background. So this tumor mutation burden
is I think the first FDA approved biomarkers in cancer. So to quantify which patient can receive
this immune checkpoint blockade therapies. So FDA gives a threshold of 10. So that for patients,
(08:20):
if they have tumor mutation burden of 10 mutations per megabase on their genome,
they will be qualified to use this immune checkpoint blockade therapy.
However, later research found that this is a suboptimal biomarker because it will miss
a lot of real responders. That is there are a lot of, in our study we found that
(08:45):
maybe there are more than 60 % of the responders they have a tumor mutation
burden less than 10. So this biomarker is quite suboptimal and this is why we develop LORIS.
From LORIS we found that we can identify patients that are previously thought to be poor candidates
(09:06):
for immune checkpoint blockade therapies but they are actually the responders. So we hope
this tool can help clinicians to select more patients that can benefit from this therapy.
Oliver Bogler (09:20):
The model that you developed is based on historical data,
right? On retrospective data that you had. And I think you mentioned in the paper
that you need now as the next step or one of the next steps,
a prospective study to really test it. So tell us about that. What would that look like?
Tiangen Chang (09:39):
Yeah, so we have basically in clinic we have two types of data. One is
retrospective data, one is prospective data. Retrospective data that is collected before
that is already there. So you can actually you can do cheat, right? Because you already see the
result so you can modify your classifier so to fit the result so you can cheat. However,
(10:05):
for the prospective data you cannot do that because the data is still not there. So you set up
a new clinical trial and then you have the model first. Then you collect data to see if your model
really works or not. So prospective data is really important in computational science. So that can
evaluate if biomarker or findings really true or not because in this case nobody can do the cheat.
Oliver Bogler (10:32):
Interesting. So Yingying, a different question. I don't know,
when I looked at the paper, I read the paper and looked at the model,
it struck me as surprising that only six pieces of data can be so powerful
in that prediction. And that may be because I lack mathematical intuition, but were you
surprised by that? We always talk about cancer is so complicated, there's so many things happening.
Yingying Cao (10:55):
Yes, I think it's quite surprising because intuitively,
in the machine learning study, the more features should have better performance. But currently,
we found that these six features with the current data have the best performance.
I think it's not intuitive, it's surprising. The reason is that,
(11:21):
firstly, usually the clinical data is quite noisy. And when the data is quite noisy,
it doesn't mean that the more complex model or more features we have for the better results.
And another thing is that it doesn't necessarily mean because of the noise of the data and the
(11:45):
more features can cause overfitting. This is also a major, I think the major benefit of
our model that it can be generalized to test the data and the data set from other cohorts.
(12:07):
So I think it can be say surprising and it's not surprising because we can explain that.
Eytan Ruppin (12:14):
I would like to pitch in just briefly. So first of all, Oliver,
the simpler the model is, it's the better and it's more believable. So
simple is an advantage in the parlance of models, complexity, machine learning,
what have you. It's called Occam's razor. So simple is good. That's one.
(12:41):
Second of all, a lot of features are strongly correlated between them. So the fact that you
see only a few gives you a lot of information, hidden information also on other features.
The question is to pick the ones that are most informative on the
(13:03):
state of the system. And that is what very nicely Tiangen and Yingying have done.
But thirdly, we should remember that we think that if further tested and validated going forward,
this could be valuable. That's our hope, of course, our belief, our belief, okay? But we
(13:29):
are far from perfect. So the performance without going into technical details of measurements of
performance is 75%, 80 % accuracy, but not more. Okay? And we should remember that. And
that's a very important open challenge and we are not yet there. We are not there yet.
Oliver Bogler (13:54):
So it's a really good first step, but there's more to come.
Eytan Ruppin (13:58):
Exactly. And the way I would relate to it and even if prospectively verified and so on
is another way to help clinicians take the best decisions. It's not that you know the
futuristic whatever the AI medical robots that are now we are very far from that.
Oliver Bogler (14:21):
Right. So, Tiangen, the other thing I noticed is that this model has predictive powers
in a number of different cancers, right? It's not just for one type of cancer. And
I guess that points to a commonality of mechanism amongst these cancers in
relationship to the efficacy of the checkpoint inhibition. What are your thoughts on that?
Tiangen Chang (14:42):
Yeah, yeah, thank you for the great question. Yes. So yeah,
we, because we first of all, we collect a lot of cohorts, different cohorts, they have different
cancer types. So our first question is if we can build one model that can capture the commonality
between cancer types to get a model that predict the ICB response in different kinds of types. So
(15:08):
then that is our first model. And that as you may know, this model captures the commonality
but does not capture the specific special property of individual cancer types. So then we developed
the second LORIS model that is specifically lung cancer. We’re doing lung cancer because this is
(15:29):
a cancer type that we have the largest sample. So we want to still keep the statistic power.
Then we found that using the same strategy we can also predict the lung cancer immunotherapy
response quite well. But the model is slightly different and the lung
cancer we added one feature that is PDL1 expression. And this is a commonly used
(15:53):
feature in lung cancer. Clinicians already know that the PDL1 expression is a very
important feature that determines the immunotherapy response.
Oliver Bogler (16:05):
Right, PD -L1 is one of the checkpoints, right? The checkpoint
inhibitors try to knock down, so to speak, to allow the immune system to fire. Is that right?
Tiangen Chang (16:15):
Exactly, yeah, it is expressed on the cancer cells and stop T-cells to kill them.
Oliver Bogler (16:20):
Eytan, thinking about the LORIS work and the many other things your group does,
and you just told us we're not about to have medical robots coming into our hospitals,
what do the next years hold for your AI work and AI in general in this domain? What's coming up?
Eytan Ruppin (16:41):
Yeah, that's a fantastic question and I do not know the answer. I can tell you what
we are working on and excites me and I'm hopeful about. So we is my lab and our collaborators of
(17:02):
course. So we are working on two I am gung ho about developing methods to better match
treatments to patients. Okay? And I'm trying to do that in two main new dimensions. One of them is,
(17:24):
believe it or not, look at the good old hundred years pathology slides, you know, which are the
backbone of tumor classification and and via that some treatment recommendations and actually apply
the recent developments, buzzwords in AI and machine learning and deep learning and all that,
(17:49):
and predict the molecular characteristics of tumors directly from the pathology slides. The
gene expression, the proteomics, genomic alterations, mutations, what have you.
This is now an extremely hot field with a few fantastic labs working in the US,
(18:11):
Europe, whatever. And we are also heavily excited and invested in that. Actually,
we just published a paper in Nature Cancer today on this topic. And a
month ago in Nature Medicine, so you can see the pace, the pace of things.
(18:32):
And in this paper that we published today, we show that we can actually take the slides and bring,
predict omics data from them. And based on that, in previous work that we have done,
bring three recommendations. And we are showing validations in five different
independent cancer types and treatments. So that's one direction. Slip sliding again,
(18:58):
Oliver, if you remember Paul Simon's, I mean, “the kids” have no idea what we are talking about,
but right, the young researchers. So that's one direction, which I'm gung ho. What can
we do from the slides? And many others, of course, others doing fantastic work on it.
(19:21):
The other topic which Tiangen and Yingying are actually taking active part in
is asking what can we learn from the blood on the tumor immune microenvironment? And so
(19:41):
together with other wonderful collaborators in the NCI, we are collecting matched data
where we have from the same patients, both the blood and tumor molecular information.
We are doing what is called single cell sequencing and what have you and whatever.
And we are learning what we can predict from the blood about what's happening in
(20:07):
the tumor microenvironment. And for that, for example, predict the response to immune
checkpoint therapy. So this guy, Tiangen here, has now a paper finding with some
very exciting findings on that which we are going to submit in the next few days.
(20:29):
And you know, if I'll tell you, I'll have to kill you, dear.
Oliver Bogler (20:34):
I won't ask.
Eytan Ruppin (20:36):
So anyway, so that's the two main directions where, you know,
I'm excited about and I believe that we can deliver meaningful advancements in the next
few years. And here, this is recorded so you can call me accountable, right?
Oliver Bogler (21:00):
Yeah, we'll check in in a few years and see how you did.
Eytan Ruppin (21:03):
The Virtual Site Visit.
Oliver Bogler (21:05):
Exactly. Those are exciting areas. Sounds fantastic. Tiangen, I know I can't ask you
specifics about the paper you're about to submit. It sounds really interesting. But
is it building on LORIS? Is it a natural? No, it's something completely different. Tell us.
Tiangen Chang (21:23):
Okay, yeah. So yeah, currently I actually, yeah, I continue working on predicting
and understanding the response to immunotherapy. But now I'm more focusing on studying the tumor
microenvironment. So I do not limit myself to predict the immune checkpoint blockade response
(21:44):
only from clinical data. So for the paper Eytan just mentioned, so yeah, we use more advanced
techniques such as single-cell RNA-seq so we can look at individual immune cells and tumor
cells in the tumor microenvironment and also in the blood. So we use a similar strategy. As
(22:06):
Eytan said, we computational scientists, we collect data from public databases.
So we collect a huge cohorts of data. To our knowledge, it is the largest head and neck
cancer. This is a single cancer type. And we found the B-cell signature predict the immune checkpoint
(22:28):
blockade response quite well. And it only, not only works in the tumor microenvironment,
but also works in the blood. So simply from get the blood of our patients,
then do a single cell or even simpler, do the flow cytometry. We can predict
the response probability of this patient to immune checkpoint blockade therapies.
Oliver Bogler (22:51):
Right, so this is a kind of a new phase where you're not using clinical data,
you're using more biological data that are perhaps also harder to acquire than routinely
perhaps collected in the clinic. But is there a future where the two might converge and allow you
to have a more comprehensive prediction model? I mean, Eytan mentioned that the LORIS model
has an accuracy of approximately 80%. So will you get to higher numbers with this approach?
Tiangen Chang (23:17):
Yes, we hope so. Yes, one of the advantages of machine learning and this
artificial intelligence is to integrate data from different modalities, from
the clinical data, from the molecular data, and to integrate all this data. So we hope
we can increase the predictive power to lead us to a more accurate prediction.
Eytan Ruppin (23:43):
So, yeah, so I want to mention in that aspect, you know, it's very hard,
Oliver, there is little molecular analysis of patients' data yet because of the costs and
because it's not part of routine practice. So, where we hope that it could be a game changer is
(24:04):
these technologies where we look at the slides and infer the molecular data and we can do it
pretty well now. That can potentially enable us to contribute to democratizing medicine.
So you know, even within our nation, there's huge differences between different regions and
(24:26):
so on and so on. It is our hope that if further developed, this could be a very important factor
and I truly believe we can do it. I am actually, together with Ken Aldape, our chief of pathology
and my good friend and collaborator, we connected a few foundations that do work in Africa. And
(24:50):
we are actually engaged almost every week. We spend time meeting with at least one or two groups
enlisting them to get data from Africa, deploy our methods. There's a whole host of problems and
challenges. It's a topic for another discussion and interview. But we are persistent people.
(25:16):
So actually, I grew up as a kid in East Africa. And I actually don't test my Swahili,
but I actually have a deep connection and a moral commitment is a strong word,
but whatever interests. So I'm very excited to see how it will unfold. And the people we speak to,
(25:43):
the medical oncologists and the pathologists in Africa are really excited about that. I
hope we can look forward and make a difference. It's early days. Yeah.
Oliver Bogler (25:53):
But that's exciting. So the humble H&E slide may put the
sequencing companies out of business.
Eytan Ruppin (25:59):
Never. That's my plan. That's my plan. You know,
the revenge of a lowly federal clerk.
Oliver Bogler (26:10):
But exciting, as you say, it will make that kind of information available
in many more parts of the world where the kind of deep sequencing is still out of reach. That's
exciting. And then I understand those molecular data inferred from the H&E slide could then be
combined into predictive models that you're building. Yingying, how does your next piece
(26:32):
of work, your next project fit into this scheme now that LORIS has been published?
Yingying Cao (26:37):
As you know, that LORIS used the features from the clinical routine collected data,
and they cannot explain more details in the mechanism of how the immunotherapy
response in each patient is and how the tumor in the human body,
(27:00):
the tumor microenvironment react to the tumor cells. So I'm more interested in the details of
the interaction of tumor microenvironment and the tumor cells. So currently I use the single-cell
RNA-eq data and the spatial transcription data to really understand the tumor microenvironment.
Oliver Bogler (27:27):
Fantastic. I'm so excited to hear about all the things you're all
doing and I can't wait to see what amazing advances you make. We're going to take a
quick break and when we come back, we will be talking about career paths with our guests.
[music]
Cancer researchers, are you drowning in a sea of scientific papers? Wish there was a way to
(27:49):
cut through the noise and find the research that truly matters to you? Look no further than NanCI,
the revolutionary new AI-powered app from the National Cancer Institute.
NanCI isn't just another research publication app. It's your personal research assistant,
designed to understand your unique interests and connect you with the most relevant papers, people,
and events in the field. Imagine being able to chat directly with research papers, ask questions,
(28:15):
and get instant answers. That's the power of NanCI's AI-driven "chat with paper" feature.
But NanCI is more than just a conversation starter. It's a networking tool, helping
you discover new connections and build your professional network. It's a learning tool,
recommending papers you might not have found on your own and helping
you quickly get up to speed in new research areas. Simply declare an
(28:35):
interest by bookmarking papers on a topic into a folder and NanCI helps you stay up to date.
Whether you're a seasoned researcher or early in your career, NanCI can
help you take your work to the next level. By harnessing the power of AI,
NanCI is transforming the way cancer research is done.
Download NanCI for iOS today from the Apple App
Store! That is N-a-n-C-I. Experience the future of research for yourself.
(29:02):
[music]Oliver Bogler: All right, we're back. Eytan,
you did an MD and a PhD in computer science, an unusual combination. How did that happen?
Eytan Ruppin (29:12):
Serendipitously… or whatever you say it in English, right? So actually my
favorite talk to young fellows in the NCI is the title is ‘how not to build a scientific career’.
All the mistakes I made and there are so many Oliver, there you'll be. So yeah, I started late,
(29:39):
I studied medicine. And then at the age of 30, I finished my studies, finished my internship
and I came to the conclusion that I do not want to be a practicing physician. And then
what will I do, right? And so I was really thinking, I was quite lost, to be honest.
(30:08):
And then I came upon a book, because you asked us to think about the book
or movie or something that influenced our lives. And that is a book that changed
my life. So the book is called Bach, Escher, Gödel by Douglas Hofstadter.
Oliver Bogler (30:27):
Yes, I've read it, yeah.
Eytan Ruppin (30:29):
Yeah, okay. Did it change your life?
Oliver Bogler (30:30):
It didn't change my life. Yeah, but long time ago.
Eytan Ruppin (30:32):
So that's an amazing book about the deep connections between computer science
to music and art and the concept of recursion, but brilliantly written. And I decided that I
want to become a computer scientist. So at my time, you know, all the dinosaurs
(30:52):
roamed the earth and you remember them, they were big and green and frightening,
right? There wasn't an MD, PhD or something in my country. So I had to do a first, second and
third degree in computer science in parallel to my residency in psychiatry. So I didn't sleep
(31:17):
for about five or six years. And luckily I worked hard and did it in record time. And
then I went to a post-doc and a residency in the US and the rest is history.
Oliver Bogler (31:35):
Interesting. Yeah, you founded the Cancer Data Science Lab at NCI, right?
Eytan Ruppin (31:40):
Correct.
Oliver Bogler (31:41):
And we'll get back to that in a moment. But first I want to ask Yingying,
what got you into science in the first place? What was that moment
that made you say, I want to do computer science?
Yingying Cao (31:53):
During my bachelor study, I studied preventive medicine and
I even don't know what bioinformatics is after I graduate from bachelor degree. I was enrolled
into a bioinformatics group during my master degree study. And I found it fascinating that
(32:15):
because in preventive medicine study, we focus on the population medicine. We focus on how to
make some strategies to prevent the disease. For example, the change of the living style,
(32:35):
just to make some strategies to do on the population level. So we don't do it on the
individual level. And during my bioinformatics study in my master degree, I found that each
human body can be quantitatively, can be a quantitative object. So each molecules in the
(32:59):
human body and even the abundance of different cell types can represent your immune level.
So as I found it's quite interesting to treat each human body as a quantitative object and
to see how the human body can fight off the infectious disease. The math degree I studied
(33:24):
the virus infection, infectious disease. Actually the cancer research and the infectious disease
they share very common things, especially in the immune immunity. So I found that immunity
is a very fascinating subject. And then I entered NCI to study the cancer research,
(33:50):
the cancer immunotherapy. And I found it, yeah, I think every step is unexpected. When I studied,
I found it interesting. And so I don't know what the next step is, but I found I'm more and more
interested in the scientific research. And I want to build my career on the scientific research
(34:18):
to be in the future maybe a mentor because during my scientific past, I meet a lot of
very good mentors. I want to transfer what they teach me to the trainees. And also,
I also have a lot of ideas even now that, but I cannot do it alone. So I want to build a
(34:44):
scientific team to include different talents and to collaborate and to...
find the truth about immunity, the cancer research, and even some infectious disease.
Oliver Bogler (35:00):
Thank you for sharing that. So you're kind
of immunity first and then you pick up the tools you need as you go along. Tiangen,
how about you? What got you interested in science and computer science?
Tiangen Chang (35:13):
Sure I think yes it's a great question and the answer for me is
simple I think it's largely due to personality so when I was young I can spend a lot of time
sit down to work on a problem for a very long time and I was not very good at playing games
(35:33):
or socialization with people and I enjoyed reading walking in the nature or exploring
science. So I don't like doing repetitive work but I found a great joy in spending a lot of
time thinking about a problem and eventually finding the answer. So that moment of discovery
(35:58):
makes me so happy and I enjoy that feeling immensely.
So actually I started my PhD career in plant science before that I have a master bachelor
degree and at that time I knew little about biology. So I didn't find much
difference between animal science and plant science. My PhD supervisor was wonderful.
(36:22):
He taught me two very important lessons that I still benefit from it. So firstly,
you need to trust yourself and focus on important questions. Do not just follow others. Secondly,
focus on translationally focused research. I'm not saying that basic research is not good,
actually basic research is extremely important but you know every supervisor
(36:47):
has their own style and this style will influence greatly on their trainees.
At 2021 I decided to study cancer this year one of my uncles diagnosed with cancer and before
that two of my grandfathers also had died of this disease. So I decided to devote myself
(37:13):
in fighting cancer. And I always feel very fortunate to have the opportunity to get into
this field. So I feel very thankful for Eytan. And even though I don't know why he chose me,
because at that time I was a completely clueless guy. Maybe that day he was drunk,
(37:34):
but that was a really beautiful mistake. I also want to thank Yingying, she joined
Eytan’s lab a little bit earlier than me and he introduced Eytan to me. So this is my story.
Oliver Bogler (37:48):
Thank you. Thank you for sharing that. I mean, this is really interesting to me.
So all three of you really started in biology and in medicine in different parts of that ecosystem,
but came to data science more out of an interest or a recognition that that tool set and that way
of approaching a problem could be interesting and helpful. Right? Is that, did I get that right?
Eytan Ruppin (38:13):
Yes, but not exactly in my case. So I...
Oliver Bogler (38:15):
Okay. I mean, you got lost and then you found yourself.
Eytan Ruppin (38:19):
I'm still drunk according to Tiangen.
Oliver Bogler (38:21):
Ha ha ha.
Eytan Ruppin (38:22):
So no, so actually, you know, my career has been so tumultuous. So I actually,
because I did my residency in psychiatry, I was naturally interested in understanding
what's happening in the brain, you know, in neurological and psychiatric disorders.
(38:42):
So I actually spent my first 15 years of my career as a computational neuroscientist building models,
neural network models, understanding, trying to understand what happens in Alzheimer's,
schizophrenia in the brain. And after, I don't know, I was already a professor of
medicine and computer science accomplished blah, I came to realize that I will not
(39:11):
understand it. I don't believe that in my lifetime, anyone else will understand it.
So I decided to take the bold step and reinvent
myself completely and move to a much easier problem, how to solve cancer.
Oliver Bogler (39:27):
Fair enough.
Eytan Ruppin (39:28):
That's true, that's true.
Oliver Bogler (39:29):
Yeah? And how's it going?
Eytan Ruppin (39:31):
Not so bad, I mean, surprisingly good. I mean, you know, our ability to understand
what's going in cancer and to come up with better treatments is advancing, I think, objectively or
not objectively at a much more impressive pace than what we understand about the brain how
(39:56):
the brain works. So I'm very, very happy that I finally made one good choice in my career. Yeah.
Oliver Bogler (40:09):
I'm sure you made many, but we're very excited that you chose cancer
because we need incredible people like you and your team. So I wonder as my final question,
what is your advice to people who are listening, who might be primarily interested in biology,
medicine, cancer, and be thinking about, well, how do I continue to do that in the current AI
(40:31):
revolution? Or conversely, coming from the computer science side and thinking about how
do I come into the world of cancer? I want to do something really important and meaningful
in my life. What's your advice to people, young people thinking about these questions?
Eytan Ruppin (40:47):
That's a wonderful question. I'll be happy to give you my take,
but I want first to give a chance to my fellows to answer it, and then I can take it from there.
Oliver Bogler (40:57):
Please.
Tiangen Chang (40:59):
Yeah, I think yeah from my experience you need to first to choose
a important question, so research question, then you just focus on this question. I still
remember that three years ago when I planned to join Eytan’s lab I prepared a research
plan so I just want to appeal to attract Eytan’s to see how beautiful this plan is,
(41:25):
how thoughtful I was and so he chose me. But yeah, now when I look back,
it's very naive. As Eytan said, all is very good. Only problem is that it cannot be realized.
But I found that still the question is still there. So at that time,
I just want to work on the cancer immunotherapy and the tumor microenvironment. And I'm still
(41:51):
working on it. And I will work on it for the next five, 10 years.
So I think the scientific question is really important, but the method will change a lot
because with the advancement of measurement and the computational tools. So for me, I never put
(42:12):
a major of my attention on the technical things. I put more attention on the scientific questions.
Oliver Bogler (42:18):
Yingying, what's your perspective?
Yingying Cao (42:20):
Yeah, I agree with Tiangen that I think the scientific question
in your mind that the puzzle you, that motivates you is very important. If you
don't have that question in your mind that you want to solve it, it's difficult to continue
because you will meet a lot of challenges and difficulties during the scientific research. If
(42:46):
there is no scientific question that h as you and the kind of mud can make you sleepless,
it's difficult to continue. So I think it's very important to find a real interesting question.
For me, it is really fascinating the tumor micro environment and, I'm very lucky. I think this area
(43:13):
is a very lucky area for bioinformaticians because more and more advanced techniques
and currently the spatial techniques can give you the information, the spatial information
of the molecules’ and the cells’ locations and even the distance between different cells. And
(43:36):
it's unbelievable maybe 10 years ago and even people are interested in it. It is difficult to
study that in such details. And also we can do it in single cell level and the spatial level. And
so I think more and more advanced techniques can help us to understand our questions in our mind.
(44:01):
And there is one part I don't agree with Tiangen is that Tiangen don't think the
techniques develop algorithms or techniques are important. I think the algorithms and the
techniques are very important. And the techniques are very important,
(44:21):
but the algorithms are even more important because we need to understand more reliable to interpret
the data. If the tool or the algorithms we rely on is not reliable and the interpretation of the
data in the prediction of the biological system is wrong. So I think that if I develop a team in
(44:47):
the future, I will focus on the algorithm development and also the biological part.
Oliver Bogler (44:55):
Thank you so much. Thank you. Eytan,
what do you say to people thinking about a career in your area?
Eytan Ruppin (45:03):
Yeah, so I will answer at the higher level. Okay. You know,
at my rank, we are clueless. We only talk at the higher level, right?
Oliver Bogler (45:15):
Of course.
Eytan Ruppin (45:16):
So seriously, you know, only choose this career if you love it. If you're excited by
science, if you're excited about the cause, you know, listen to your heart. Like in any decision
we make in life. You can think from here to the moon and back but listen to your heart.
(45:40):
Second of all, be a nice person. It is so important for your success. Science
today is a team work, both in running your own lab
and creating a good atmosphere. I always say I only hire and work with nice people,
(46:05):
okay, because I'm not paid enough as a lowly federal clerk to work with not nice people. No.
And also, with the collaborators. It's so much emotional IQ, how it's called, you know,
in leading, not less important than every other aspect. So be a nice person, an honest person,
(46:30):
a good collaborator. You know, we are at the data science lab when I was hired by
Tom Misteli. You know, he asked me, what's your vision, Eytan? And I said, you know, we will be
a center. We will help people. And I believe we do. We are not only doing our own research,
(46:50):
of course we are, but we are helping a lot of other people and other people are helping us.
So that's the second thing. First one was love. The second one was goodness and kindness. And
the third one is bravery. Be brave, okay. Pursue important questions. Don't strive,
(47:19):
especially it's a tendency, I don't know, in young PIs or whatever, to publish a lot. That's stupid.
Try to do something meaningful. And that is especially true in CCR at the NCI,
okay? Our mandate, explicit mandate, as you know, Oliver, is to pursue high risk, as much as we can
(47:43):
projects and be persistent and relentless and try to solve them, you know, on a long range.
So just to close circle, I told you that today we published a paper and so on. This paper actually,
without going into the details unless you ask me, closes 12 years of work,
(48:07):
of a continuous work in my lab on a specific challenge. We started in
2012. Our first paper was published in Cell, so not in a trivial zone,
in 2014. Since then, we have been working tirelessly to try and improve that.
(48:28):
Okay, so the starting point was published in Cell, okay, but wasn't good enough. So luckily,
I always say to my fellows, we are so bad that we can always improve. Okay?
So because the challenges are so big, okay, so 12 years of research I was just telling
(48:50):
my wife this morning, you know, I will not mention her name because she will kill me,
that you know we published this paper and it's 12 years of the fruits of 12 years of
work. So that, you know, that makes me feel take a pause and think about it.
Oliver Bogler (49:13):
Yeah, yeah, yeah, fantastic. Well,
thank you all for that advice. Really, really good and thoughtful.
[music]
So it's time for our segment now that we call your turn,
right? Because it's a chance for our listeners to send in a recommendation that they would like to
(49:33):
share. If you're listening, then you're invited to take your turn. Send us a tip for a book,
a video, or a podcast, or something that you found interesting or inspirational or amusing.
You can send those to us at NCIICC@nih.gov. Record a voice memo,
send it along, and we may just play it in an upcoming episode. But I'd
like to invite our guests to take their turn. Let's start with you, Yingying.
Yingying Cao (49:55):
I will introduce a very interesting podcast. Its name is Intelligence Squared US. I
don't know whether you're familiar with it. I found it very interesting. It's about the
debate on the important topics, including some science and technology and the society
and the culture. And I think the different topics, as you can imagine in your life.
(50:23):
I'm very interested in listening to the debates. I found for one topic when they raise that topic,
I have some opinions in my mind. But after I listen to their debates from different part,
I change my mind sometimes and the topic become more clear to me.
(50:49):
I think this is very familiar with the scientific research. Yeah, this is what
scientific research make me interested because I always want to solve puzzles and make everything
in my life more clear. So for example, if I meet something or some challenges in my life,
(51:16):
I want to make it I want to understand it and make it clear. And the Intelligent Squared
US is a debate that can make things clear. And it make you think twice
and always step back and think twice, step back and think twice and make this clear.
And I think it's a very interesting podcast and people can learn a lot.
Oliver Bogler (51:39):
Thank you. Thank you for that tip. We'll put a link
in the show notes. Tiangen, what have you got for our listeners?
Tiangen Chang (51:44):
I would like to recommend a book and the title of the book is World Views,
an introduction to the History and Philosophy of Science. The author of the book is Richard
DeWitt and this book is a great read that takes you on a journey through
the evolution of our understanding of the world from ancient times to the modern day.
(52:06):
This book really blew my mind by showing how different world views can lead to
completely different understanding of the same universe. And after reading this book,
I have had a feeling that even now most of our scientific findings could still be wrong. But
they are meaningful because they represent the steps towards a better understanding of our
(52:29):
world in the future. Just like a famous saying that all models are wrong, but some are useful.
So finally, I want to highlight that what is really cool about this book. So this
book just breaks down very complex ideas in a way that is extremely easy to follow.
So it is perfect for anyone who is curious about our world, science and its history.
Oliver Bogler (52:54):
Thank you. That sounds really great. I might just pick
up a copy. That sounds really interesting. Eytan, your turn.
Eytan Ruppin (53:02):
Yeah, so I just work guys. I don't read anything. I don't listen. I have
no time. I don't understand. I'm a sucker of popular science and science fiction. So,
you know, very large genre and I don't know if you know, Isaac Asimov, who is,
(53:24):
and so on, actually, in addition to being a fantastic science fiction writer and,
you know, the series of the robots and, you know, other series, he also wrote fantastic
popular science books. Really fantastic. Once when I had time, I would read them.
(53:51):
What I do now is I subscribe to New Scientists. New Scientist is in my mind a fantastic popular
science. I read a lot. I'm an aficionado of physics and in another world I would like to
be a physicist. Somehow it didn't, wasn't in the cards for me in this life, but maybe in another
(54:16):
life because I'm so fascinated. And our youngest son studies physics, so I live vicariously by him.
And my favorite thing is to watch different physics, popular physics, you know,
shows in series with him whenever he comes home. And then he explains to his stupid dad, you know,
(54:41):
very complex concepts. And he's very patient with me. And I'm so happy. I feel so proud.
I'm so happy to be stupid with him. You see what I mean? And I live vicariously. So I recommend
(55:01):
New Scientist. I'm not a show shareholder. I have no conflict of interest. I love popular science.
Oliver Bogler (55:13):
I think that's a really good recommendation. Thank you very much. Well,
thank you all three of you for spending time with us and coming on the pod and
sharing your insights and about your exciting work. I wish you all nothing
but the most tremendous success. You're really doing such important work. And so thank you.
Eytan Ruppin (55:32):
I want to thank you really very much for having us. And I'm not just
being polite. You know, I come from the Middle East, I'm not polite. By definition,
right? But I always talk sincerely. So I think it was very enjoyable,
I hope my colleagues, Yingying and Tiangen, shared the same thing.
(55:56):
[music]
Oliver Bogler (55:56):
That’s all we have time for on today’s episode of Inside Cancer
Careers! Thank you for joining us and thank you to our guests.
We want to hear from you – your stories, your ideas and your feedback are welcome.
And you are invited to take your turn and make a recommendation
to share with our listeners. You can reach us at NCIICC@nih.gov.
(56:17):
Inside Cancer Careers is a collaboration between NCI’s Office of Communications and Public Liaison
and the Center for Cancer Training. It is produced by Angela Jones and Astrid Masfar.
Join us every first and third Thursday of the month wherever
you listen – subscribe so you won’t miss an episode.
If you have questions about cancer or comments about this podcast,
(56:41):
you can email us at NCIinfo@nih.gov or call us at 800-422-6237. And
please be sure to mention Inside Cancer Careers in your query.
We are a production of the U.S. Department of Health and Human Services,
National Institutes of Health, National Cancer Institute. Thanks for listening.
Popular Podcasts
Crime Junkie
Does hearing about a true crime case always leave you scouring the internet for the truth behind the story? Dive into your next mystery with Crime Junkie. Every Monday, join your host Ashley Flowers as she unravels all the details of infamous and underreported true crime cases with her best friend Brit Prawat. From cold cases to missing persons and heroes in our community who seek justice, Crime Junkie is your destination for theories and stories you won’t hear anywhere else. Whether you're a seasoned true crime enthusiast or new to the genre, you'll find yourself on the edge of your seat awaiting a new episode every Monday. If you can never get enough true crime... Congratulations, you’ve found your people. Follow to join a community of Crime Junkies! Crime Junkie is presented by audiochuck Media Company.
24/7 News: The Latest
The latest news in 4 minutes updated every hour, every day.
Stuff You Should Know
If you've ever wanted to know about champagne, satanism, the Stonewall Uprising, chaos theory, LSD, El Nino, true crime and Rosa Parks, then look no further. Josh and Chuck have you covered.