Revolutionizing Cancer Research: Navigating the AI Landscape

Episode Transcript
Available transcripts are automatically generated. Complete accuracy is not guaranteed.
OLIVER BOGLER (00:03):
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 two members of the NCI's Center for Strategic Scientific Initiatives,
which strategizes and coordinates many of NCI's interdisciplinary

(00:27):
cancer research programs and has taken an interest in how
artificial intelligence will transform cancer research, prevention and care.
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. And of course, we're always glad to get your
feedback on what you hear and suggestions on what you might like us to cover. The
show's email is NCIICC@nih.gov.So with us is Dr. Juli Klemm,

(00:55):
who directs the Informatics Technology for Cancer Research program and chairs
NCI's Artificial Intelligence Working Group among many other roles at CSSI. Welcome, Juli.

JULI KLEMM (01:05):
Thanks Oliver, great to be here.

Also with us is Dr. Sean Hanlon, who serves as the Acting Deputy
Director of the Center for Strategic Scientific Initiatives and provides
scientific leadership to the 4D Nucleome Program, NCI's Human Tumor Atlas Network,
and several NIH committees and working groups. Welcome, Sean.

SEAN HANLON (01:23):
Thanks, Oliver. It's super exciting to be here.

So Juli, let me start with you. As the director of the NCI Informatics
Technology for Cancer Research program, the ITCR, could you describe how this
program is integrating AI and machine learning tools into cancer research?

JULI KLEMM (01:41):
Yeah, sure. So this is a program that the NCI has been running for over 10 years
to spur the development of novel informatics tools and resources that can be broadly applied
across various types of cancer research across the research continuum. And one thing that's really
exciting about this program is that the types of technologies that investigators are developing

(02:08):
and submit through the program really parallel the advances in cancer research technologies.
And so, you know, for the last several years, many of the applications we've received through
the ITCR have been incorporating advanced machine learning and deep learning methods. And that has

(02:30):
just skyrocketed in the past several years, especially with now the availability of large
language models and other types of generative AI.We're really seeing that the majority of
applications coming in through the program are leveraging these technologies to advance
all kinds of aspects of cancer research. That includes discovery biology, clinical trials,

(02:57):
epidemiology, all kinds of computer vision. It really spans the cancer research continuum.

That’s really exciting. Some of the applications I think that
I'm familiar with are, for example, image analysis in clinical research,
looking at MRIs or other kinds of images. I heard lot recently about modeling molecules
and things. What other examples of how AI is being used can you tell us about?

JULI KLEMM (03:24):
Yeah, and you've touched upon where I think AI has had the most impact in cancer
research is in image analysis and particularly in whole slide images where these days, slide
images are now digitized and the digitization of these slide images has made it possible to
run these large scale analyses and train large models to find sub-visual features that can be

(03:50):
correlated with drug response or outcomes or cancer subtyping and so forth. So that really
is where the biggest impact has been to date.But now we're also seeing impacts in areas such as
clinical informatics, where these large language models can be applied to extract information from,

(04:13):
say, unstructured text and find ways to correlate treatments with outcomes, especially in rare
cancers where say no one institution might have a large amount of data. So natural language
processing and text extraction has had, there's been a significant impact on those in that field.

(04:38):
But also, in applying these, applying these technologies to understand basic mechanisms of
cancer biology. And for example, understanding being able to predict types of cancer that
are going to be most responsive to certain drug types or understanding interactions in the tumor

(05:04):
microenvironment. So it's really percolating through all aspects of cancer research.

Fascinating. Sean, I mean, just keeping up with AI, which in itself covers a
multitude of different things, feels crazy. Being involved in trying to see how the NCI
can best employ these kinds of technologies and harness them, it must feel like riding a tiger.

Yeah, I think that's a great analogy there. Yeah, I was thinking about this, you know,
it does sometimes feel like it's a hammer in search of a nail, right? And the people are so
excited by these technologies, they're trying to apply them to every problem. And so, yeah,

(05:52):
I think trying to make sure we're applying them where they're the most useful is really important.
So one area that I've seen some work in recently that I think is really exciting in the Human Tumor
Atlas Network, which looks at a lot of spatial transcriptomics and spatial proteomics data

(06:14):
is in cell segmentation or other ways to avoid this. This has been a really challenging step in
analyzing spatial omics data sets, just finding where the individual cells are in the images.
And so now people have been applying AI approaches to overcome this real

(06:35):
technological analytical challenge.So I think that those types of examples,
or even trying to do it in a new way. So now some people are saying, let's not even bother
doing this segmentation. Let's see if I can just say, based on just the transcriptomic images,
I can define the cell without bothering with the segmentation step. So I think when people have

(06:59):
identified, you know, real problems that can't be addressed with other tools, I think that's
a great place that, you know, AI might be might be able to really move fields forward.

That sounds really interesting. I want to dig a little bit deeper on that Human
Tumor Atlas Network work. So the goal then ultimately is to have a sort of high fidelity
three dimensional representation of a tumor and its micro environment, right? And what's going
on there, not just in terms of the cell types, but also in terms of their transcriptional activities.

Yeah, and what's referred to as recurrent cellular neighborhoods. we say,
I see a mix of different types of immune cells and cancer cells in this area. And
we see those patterns repeat across multiple different patient samples. And that can tell
you something. sometimes there's differences. If a tumor, maybe all the immune cells are on

(07:53):
the outside of the tumor versus one where you can see the immune cells infiltrating the tumor
more that might be predictive of different responsiveness to say, immunotherapies.

Sounds amazing. I mean, it's not an undifferentiated lump. There's a lot
going on. And we've known that for a while, but now we're beginning to look into it. Yeah, yeah,
fantastic. So I wonder, know, NCI both funds research, which may use AI, and it also sees
itself as a sort of organizing entity to make tools available, data available and to organize

(08:29):
sort of larger networks like the one we were just talking about. I wonder what's
the balance in your minds and what are the strategic priorities for NCI say in the next
five years regarding AI? Maybe that's not a fair question, but I asked it anyway.

I mean, part of this is my personal interest. I feel like making the data available is
a priority. We've paid for this data already. And making that usable and accessible to the broader
research community really leverages investment and gives other people opportunity to build on

(09:07):
findings. Just going back to the Human Tumor Atlas Network, again, an example of this:
we've been trying to get people more access to these and we've been hosting what we call data
jamborees where we actually bring in groups of people to form small teams and work on projects

(09:27):
that are based on reusing the HTAN datasets to try and identify new hypotheses or build new tools. So
really getting people hands on working on this.And I think, going back to the bigger picture
of priorities, think engaging with communities and getting a wider set of people to use it. We

(09:52):
can get new perspectives and it's just a way to, think, democratize access to
data and democratize who's being able to participate in cancer research.

JULI KLEMM (10:05):
Building on what Sean was saying about really providing some of the foundational
resources needed to make use of this technology. I also, I think from the NCI perspective, helping
define how models can be shared in a way that they can be maximally reused. So helping people,

(10:29):
helping define what information you need, to make available to describe a model,
such as what data was it trained on, how has it been validated, what is its intended use,
how should you not use it, and so forth. And so actually within the ITCR program,
our most recent funding opportunities incorporated guidelines for how to share

(10:53):
models so that we can maximize the reuse of these technologies.

Yeah, I mean, that sounds like being a good
steward of the taxpayers' funds and also accelerating cancer research,

JULI KLEMM (11:08):
Exactly, exactly. So that, you know, that we can build on these investments
and continue to, you know, elaborate on models that are showing promise.
You know, I think also, you know, because this technology is evolving so quickly,
I think we all know, even in our personal lives, how, you know, AI tools are available,

(11:29):
you know, to apply in all aspects of our lives. But I think especially in healthcare,
we need to make sure that they are having the intended consequences. And so the idea
of really supporting the evaluation of these models in real world settings,
I think is very important because there is huge potential in advancing aspects

(11:54):
of healthcare and clinical decisions. But we need to make sure that it is having the
intended consequences and supporting research to understand that I think is also very important.

I’m curious from a career angle as you look across your portfolios and your programs,
is it still the domain very much of the computational scientists or are there also
regular biologists like myself for example who are daring to venture into this area?

JULI KLEMM (12:21):
I mean, I think this gets into the considerations of team science
and multidisciplinary science. Especially speaking from ITCR,
most of the teams developing tools for cancer research, these informatics tools,
are interdisciplinary teams of people maybe classically trained in computer science, working

(12:45):
closely with oncologists and cancer biologists.However, to your point, Oliver, I think some
of these new tools can be leveraged more directly by individuals who maybe weren't
originally trained in computer science, but can make use of generative AI to extract information

(13:07):
from large corpuses of text, defined patterns in data. So I think that is evolving as well.

Yeah, I guess I would just echo what Juli was saying. And I think
most of the programs that I'm involved in are very transdisciplinary. it is the computer scientists
working with the cancer biologists, the oncologists. So I think one important thing
is for people to at least have familiarity with the different fields and learn the

(13:41):
language in order to be able to talk effectively, collaborate effectively with people in different
disciplines. And I think that's something that people can start doing as trainees.
And I think it would be really valuable for their careers and something, like you said,
where maybe we’re riding the tiger at this point, but things are changing and you don't

(14:03):
know exactly where things are going to end up in five years. But those collaborative tools and
experience and expertise will be valuable I think regardless of where we end up in the AI space.

So you recently launched a seminar series the Cancer AI Conversations,
which I think have been very interesting. Can you tell us
a little bit about what those are and what the thought behind that series is?

JULI KLEMM (14:31):
Sean and I, together with another colleague, co-chair the NCI
Artificial Intelligence Working Group. And this working group started before the emergence of
large language models and generative AI. And you know, when we first started, you know,
had more traditional workshops that NCI organizes, which is, you know, planned out for a few months,

(14:55):
think carefully about what experts to bring in, carefully outline an agenda and then have a very
in-depth conversation about state of the science.What we found in the last few years was this field
is moving too quickly to use our usual approaches to exploring the state of
the science and convening scientists. And we really realized we needed more nimble

(15:19):
approaches to explore these emerging areas of AI.So we had this idea of rather than having these,
you know, maybe a yearly workshop on a topic of interest, needed to be exploring these areas in
more real time. So we came up with this idea of Cancer AI Conversations, which would be, you know,

(15:44):
every other month discussing an emerging topic of AI, and especially as it relates to cancer
research, and to invite three experts in this field who in the field of interest that might have
complementary and different views on the topic and to really, you know, to share,

(16:05):
to give short presentations, but really to have a conversation about the topic and
involve the audience in asking questions about the topic. So to really take a more nimble real
time approach to understanding the impact of different advances in AI on cancer research.

Almost like a live podcast. What have been some of
your favorite conversations in that series so far?

JULI KLEMM (16:30):
You know, one that has me thinking, a lot of us thinking about next steps was one
of our most recent ones on the role of benchmark data sets in AI and how having
a data set that really is a ground truth information that will help validate models

(16:52):
that are aimed at addressing a certain problem.So an example is, let's go back to the example of
models for computer vision. To understand if a particular AI model is doing its,
let's say, segmentation like Sean brought up is segmenting an image correctly, you need to know,

(17:14):
what is correctness? And so a benchmark data set would be that ground truth gold standard against
which you can measure how well is this model performing in appropriately segmenting this image.
And so in that case, the ground truth might be, a pathologist or a radiologist, depending on the
image type. And you say, that is ground truth.And if the model performs as well as the

(17:41):
pathologist or radiologist in that example, you say it's performing well on the gold standard
data set. And so we've been talking about what kinds of gold standard data sets should be made
available to drive forward important areas in cancer research that could be advanced with AI.

That sounds really important though, but I wonder though:
you know, there's this old cliche about garbage in and garbage out. And Sean,
I wonder how is the NCI addressing that challenge, the challenge of imperfect data when you go out
into the real world? Let's say you develop a model and you validate it on a gold standard data set,
like Juli was saying, but then you go out into this bigger, big ugly world and the data are

(18:26):
not perfect. How do you tackle that and is AI part of the solution to that problem?

Well, mean, interesting. We did have a workshop a couple of years ago
now on AI approaches to overcome imperfect data. And so it was looking at what are the
ways that you could try and address this. So I think the community is exploring different
ways. And I think people are aware of these issues that are related to this.

(19:02):
And I think things like supporting benchmarks are important for making sure that those algorithms
that are then going out into the real world have been validated on something independent. Because I
think again, we've alluded to, this is very fast moving, lots of people are going in this field,

(19:22):
lots of algorithms are getting published. And so, if I'm going to reuse something,
how do I know how good it is? And so I want to have some of these these standard benchmarks of
ways of assessing which of the models, which one should I use for which reason. And then
having strong models then hopefully will, they'll function well with real world data.

Here on the pod, we've talked in several episodes about cancer health disparities,
which was defined as the well-documented uneven outcomes for people across America
and the underlying causes, for example, access to health care in rural areas. And I wonder,
there's been a focus on AI and cancer health disparities. And I wonder how

(20:06):
NCI is working to ensure that the AI tools and technologies and data sets
that we're talking about are developed and implemented for all Americans.

JULI KLEMM (20:16):
Yeah, that's a great question, a really important question,
Oliver. And it gets to the topic of what's referred to as generalizability in AI,
ensuring that a model that was developed at one institution, or let's say one hospital,
can perform similarly in another setting. And that ends up being really hard,

(20:40):
because there are so many aspects of the data that a model is trained on that can really impact its
performance. The population of a patient at a given hospital, the types of technologies
they're using to make measurements and so forth.And so I think there is no one answer to that

(21:04):
question, but just one area that the NCI is supporting is the idea of what's called
federated learning. And this is the idea that you can train an AI model on data from
multiple institutions without actually moving the data to a centralized location. And that's

(21:25):
particularly important on data that can't be shared for various reasons, whether that's
for patient privacy concerns or intellectual property concerns or any number of reasons.
And so the idea of federated learning is that rather than moving the data centralized to
the model, you can train the model at individual sites and aggregate the training parameters that

(21:51):
you gain at each site into a model that represents the average of all of that data it was trained on.
And that's one approach that NCI is exploring, and others outside of NCI are exploring this
approach as well, to train models that are more representative of data at many sites.

Interesting. So it's a big area. Obviously, collaboration is important. wonder,
Sean, if you could comment on how NCI is collaborating with other government agencies,
for example, and external organizations. The Department of Energy, ARPA-H, these entities
come to mind. What's sort of the big picture and where does NCI fit into that picture?

Yeah, I think that's a great question. And obviously there's lots of work
going on across the government in this space and outside of the government. You know, there's the
NCI-Department of Energy pilot projects that are one example of this. But maybe one thing
I'll talk about is Juli and I have been working with colleagues from the Department of Energy

(23:01):
and ARPA-H and the FDA in the area of using AI approaches to target undruggable cancer targets.
And I think as most people know, there's a Nobel Prize awarded in this space and in this broader
space just last year. And there's a lot of work in industry here, but there are types of targets,

(23:29):
like transcription factors or proteins with large disordered regions, and especially
these exist in pediatric cancers and other rare cancers that there's less industry interest in.
And so we hosted a workshop over the summer that brought together folks
from industry and academia and government to explore ways to help move this field

(23:54):
forward. And we are continuing to work to put out a white paper to highlight
ways in which we can work across the government to help advance this field.

So Juli, I think you already touched upon it little bit in the conversation
of federated learning. Obviously AI is all about data and since we are trying
to improve how cancer is discovered, how it's prevented, how it's treated,
patient data is central. And that raises, I think, very valid and reasonable privacy
concerns. And I wonder what the stance of the NCI is towards these sort of ethical

(24:30):
implications of AI and how we're managing these very reasonable concerns that we all have.

JULI KLEMM (24:38):
Yeah, it's a very important area and NCI, you know, takes those privacy concerns very
seriously. And I think there's that's another area I think that, you know, I think the government can
play a role in helping define how to safely share models that may have been trained on data that
includes, you know, patient information and also, by the way, making sure there's the appropriate

(25:03):
consents in place to do that training.It turns out that these AI models that
are trained on private data, in some cases, someone really clever can reverse engineer
those models and identify individuals whose data were in those models in certain cases.

That sounds scary.

JULI KLEMM (25:25):
And so we have to make sure that we treat these models
as protected information. We have these conversations, is a model software? Is
it data? It's somewhere in between, right? The model really reflects the data it was trained
on. So we have to treat them with as much sort of protection of patient privacy as we would the data

(25:47):
it was trained on. And there's some very important research underway by government, non-government,
and so forth as how to train models in a way that you can't accidentally leak private information.

Yes, sort of by prompt hacking and this kind of stuff,
right? Which we do read about in the newspaper.
All right, we're going to take a short break and when we come back, we'll talk careers.
[music]
Cancer researchers, are you drowning in a sea of scientific papers? Wish there was a way to
cut through the noise and find the research that truly matters to you? Look no further than NanCI,

(26:29):
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,

(26:50):
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

(27:10):
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.

(27:37):
[music ends]
All right, we're back. As always, my first question, how did you get started on the
path towards science that led you to your current role? Let's start with you, Juli.

JULI KLEMM (27:49):
Yeah, sure. So when I was an undergraduate,
I got interested in biological research and was fortunate enough to do undergraduate research.
And even as an undergraduate, I worked at an x-ray crystallography lab. And so that is a very
data-intensive science, and that right off the bat got me interested in data-intensive research.

(28:13):
When I did my PhD, I ended up continuing to do extra crystallography, which involved, again,
at the time, some of the most data intensive research there was. I'm old enough that
I am pre-human... my training was pre-human genome, pre-next-gen sequencing. So in a way,
I've been privileged across my career to watch incredible advances across biological research.

(28:39):
And so one of my, actually my first job out of my postdoc was at a human genome
company. This is a company that was actually sequencing the human genome.

Incyte, right?

JULI KLEMM (28:51):
That's right. I worked at Incyte Genomics. This was,
think the term bioinformatics didn't even exist at the time or maybe was new.
And so at our company, we were inventing the approaches to analyzing the deluge
of data that was being produced through these sequencing methods. And it was such
an exciting time to be in that field and to be inventing brand new ways to visualize data and

(29:17):
help people work with the data and so forth. And so I've been in data science ever since.

Fascinating. Sean, how about you? What's got you started?

Yeah, so as an undergrad, I was always planning on going to medical
school. You know, was a science, you know, a biology degree. And I never really liked the
labs because you always knew what the answer was going to be. But then as soon as I started
taking upper level lab classes, you know, I had the experience of like doing something where I

(29:51):
didn't know the answer. And I think that really intrigued me and changed my path. And I ended up,
you know, going to graduate schoolI started graduate school. I was,
I worked in yeast and I started basically just after the yeast genome was published. So that

(30:11):
really, I think has driven my interest in open science and data sharing because the availability
of the sequenced genome really change what you were able to do. You could knock out any gene,
you could clone any gene, you know, within a couple days you could create a new yeast
strain. So it was really amplified the powers of what you were able to do in molecular biology.

(30:40):
And then as a postdoc, I went to a genomics lab. This was earlier genomics, we were used
microarrays, which we printed ourselves in the basement of the building. But again,
that was powered. You could only do that because the yeast genome had been published. So you could
design primers to amplify every region of the genome and print them on a glass slide.

(31:06):
And yeah, I worked in that space. It was sort of the edge of molecular biology and a little bit
of bioinformatics. So I basically just used Excel for my analysis because yeasts have a pretty small
genome. But I've stayed in sort of in that space and I think like Juli, it's just been amazing

(31:28):
to see the technological advances from those early day microarrays to whole genome sequencing
to now single cell sequencing and spatial transcriptomics. The level of detail that we are
able to explore has just really grown so much in last 20 years and it's been amazing to watch it.

Yeah, I think we're similar. I may even be little older than the two of you
because these things sound very familiar to me. It's definitely
been an exciting few decades. I wonder then how coming out of these very sort
of foundational research areas, what drew you to the NCI and to government service?

Yeah, so I was wrapping up my postdoc in 2009. There was a bit of a financial crisis and
there weren't many jobs available for PI type jobs. And so I was exploring what I
was going to do and I ended up applying for the AAAS Science and Technology Policy Fellowship,

(32:37):
which I got and that places you in a government agency,
right? And I interviewed at a bunch of different government agencies and liked the NCI options the
best. I ended up there as a fellow and I've been there ever since for the last 15 years.

(32:58):
And I would say that the AAAS fellowship is a really wonderful opportunity for
people that are interested in exploring government agency work as an opportunity.

It’s fantastic fellowship. We'll be sure to drop a link in our show notes. Juli,
your path, you really mentioned you were in academia, you were in industry,
and now you're in government. How did that path evolve?

JULI KLEMM (33:27):
Yeah, so after working in industry for family reasons, I ended up relocating and for
a while worked in a consulting company doing consulting work in bioinformatics. And one
of our companies supported at one point an NCI program. And that was really my first exposure

(33:48):
to working with a government agency. And I have to confess that I before that did not understand
what the government did and really the inner workings of supporting academic science.
And so that consulting opportunity was my first interaction with the NIH directly.

(34:12):
And I was really intrigued by the work they were doing and the people I interacted with.
And so became very interested in a job with the government you know,
eventually applied and started working for the NCI now 18 years ago. It started in the
Center for Biomedical Informatics and Information Technology. So again, in the informatics world.

Yeah. Yeah. And we've, we've talked about that a few times on the pod,
that group, the CBIIT group. I wonder also for our listeners, I think, maybe they can imagine
what your day in the lab, was like, or sitting at the computer doing bioinformatic analysis.
But what is it like now? How much science do you get to do? Because, you know, do you, do you,

(34:56):
presumably went into discovery science because you had the science itch. Do you scratch it still?

I would say yes, definitely. I did like being in the lab and I liked carrying out
research. I didn't always have the best ideas for the big picture science questions to ask,
but I also liked really interacting with other trainees and mentoring people. So I think that

(35:32):
I still get to do a lot of that now. I get to work with investigators that are part of
these bigger programs and then I hear about the exciting science they're doing early,
right? So that's always fun to be on the cutting edge of what people are doing.
And then I also get to help connect people where I see connections between different

(35:55):
groups that are working on similar things. And I get to participate
and help organize junior investigator workshops where we're helping support
the careers of the future leaders of the cancer research community.
So I get to keep working on those aspects of science that I really enjoyed and was good

(36:20):
at. And I also get to see a much bigger breadth of science than I did when I was
working in the lab. And I think that's one of the real exciting things to me too.

JULI KLEMM (36:32):
Yeah, I do think it's that focus versus breadth is really the difference in our
day-to-day scientific opportunities. In the lab, you have the opportunity to really dive deep into
a focused problem and explore it and become a real expert in a particular domain. In our jobs,

(36:56):
we really have to take the big picture and really know, look across a field and kind of
see where there are advances being made, where there might be gaps. And I think, you know,
Sean said something that I really see as maybe the most important thing we do is we connect people
because we have that big picture. We’re able to see where there are common challenges or common

(37:20):
needs because we have that big picture. And then we can bring those people together to talk to each
other or to foster conversations around these needs and to identify, to move this field of
science forward, what is needed, and help to identify those issues and help advance them.

Yeah, that does sound really exciting and fun too. So, Sean,
you already mentioned just a moment ago the junior investigator symposia or conferences
that you hold. I wonder in general, how can our listeners who are probably early
in their career or maybe not even take advantage of some of the NCI programs

(38:02):
and resources if they're developing their own careers in this direction?

So there's an NCI supported junior investigator meeting that happens
most Augusts. I think there's some specific programs that are the leaders of that,
but it's generally open to other junior investigators. So that's one thing to do.
I mentioned the HTAN data jamborees that we host. And those are specifically targeting

(38:33):
people that are not part of HTAN and generally are made up of like 90 % are graduate students,
postdocs or staff scientists that participate in those. And we're always
looking to get new people that are interested in spatial biology to participate in that.

That's the human, I'm just gonna expand the acronym,
that's the Human Tumor Atlas Network.

Yeah, exactly. That's right.

Okay, we'll post that as well and hope that people participate.
So what advice would you give to our listeners, our early career listeners who are interested in
thinking about AI and may not, some of them may be computational scientists already,
some may not, but they're caught up in the enthusiasm for this domain. What would you

(39:21):
say to them? What skills, what areas of expertise should they focus on developing?

JULI KLEMM (39:25):
Yeah, you know, I think the reason that's such a hard question is we
are living through a revolution right now in AI, which is very exciting,
but that means we're at a time of a lot of change. So I think, you know, skills that
are gonna be future proof and that will always serve, I think young scientists well, are one,

(39:46):
communication skills, being able to communicate clearly about science. And related to that,
you know, not being afraid to ask questions and to say, don't know, so being inquisitive and humble,
I think is important. Pursuing multidisciplinary opportunities. We talked earlier about computer

(40:10):
scientists working with biologists, working with oncologists and so forth. And that also
relates to communication skills, being able to work in a multidisciplinary team, I think
is also always going to be an important skill.And finally, I think looking for good mentors.

(40:30):
Sean and I were talking earlier today with another colleague about sometimes in these
fast-moving fields, people who are older than you might not know the technology as well as
younger people. So sometimes finding peer mentors can be very valuable. But mentoring in general,
I think is always very important especially early in your career and frankly throughout your career.

Yeah, I would agree with that heartily. Sean, what advice do you have for us?

Yeah, I think again, because this is the field is moving so fast. I think you have
to be very thoughtful about, know, you want to, if you're focusing on an AI related project or
applying AI, what are you doing that's unique and that you or your group is specifically, you know,
set up to make a difference in, you know, what's good for academia because there is a lot of,

(41:25):
you know, movement in industry.So you don't want to, you know,
be you know, six months into a project and you see there's been some major advancement,
you know, on the industry side that makes what you're doing irrelevant. So think you have to be
very thoughtful about the types of problems and then questions that you're trying to address.

So I just noticed that neither of you said things like you should learn how
to program in Python or things, something like that. You've really pointed to the
durable skills that make you, I think, a great investigator. Other skills you
can pick up as you go along, right? And that's also changing so rapidly.

Yeah, yeah, I was gonna say, yeah, that you don't know. Like, if I say that,
I'm not sure that it'll be true, you know, two years from now or even six months from now, right?
[music]

Now it's time for our segment, Your Turn, where we give our listeners a chance
to send in a recommendation that they would like to share. If you're listening,
then you're invited to take your turn. Send us a tip for a book, a video, a podcast, or a talk,
or anything really that you found inspirational or amusing or interesting. You can send those to
us at NCIICC@nih.gov. Record a voice memo and send it along. We'll play it on an upcoming

(42:48):
episode. But I'd like to invite our guests to take their turn. Let's start with you, Juli.

JULI KLEMM (42:53):
Yeah, you know, two of my favorite books that are in the science realm are both by
an author, Thomas Hager. One is called The Demon Under the Microscope, and the other is called
Alchemy of Air. And they're both about fundamental advances in science that really change the world.

(43:14):
Demon Under the Microscope is about the discovery and advancement of some of the
earliest antibiotics. The Alchemy of the Air is about nitrogen fixation, which on the surface
does not sound interesting. And I will tell you, it is an absolutely fascinating history.
What I like about these books, a couple of things. One, they really intertwine

(43:37):
scientific discovery and history and show the intermingling of the historical time and the
advancement of that area of science and how that really impacted history. And second, it's about
perseverance in science. And they're both about scientists who really had to, you know, overcome

(43:59):
some difficulties to push forward something they really believed in. And so I find them,
I find both books very inspiring and fascinating. And so I recommend both of those books.

Thank you very much. Those two sound really interesting. Sean.

Yeah, I want to recommend the podcast 99 % Invisible. It's hosted
by Roman Mars, who actually is a science dropout. He was in a genetics
PhD program and left to take an unpaid internship at a public radio station and
has built up a public radio show and then into this podcast that's very popular.

(44:42):
And it's focused on the built world and the fact that you know that the title refers to
the fact that you don't recognize most things that are working well in the built world so
it tells some stories behind things that are really valuable but you might not think about
that much. So what I think one example that highlights this well is there was an episode

(45:07):
on curb cuts you know so those spots in the curb that let you roll up a suitcase or a
stroller along without having to jump up the big curb and how those originated and it was
through the advocacy of a disabled person and really wanted to have more access to the world.

Sounds really interesting. I'm a podcast addict. I will add it to my
playlist. Thank you for that recommendation.I'd like to make a recommendation as well.
I'd like to recommend the book, Mortality by Christopher Hitchens. I recently reread
this remarkable book and as a longtime admirer of Hitchens' fierce intellect and formidable debating

(45:47):
skills, I found it deeply affecting. While I don't share all of his views, his account of facing
esophageal cancer really resonated with me. What struck me most was his unflinching examination of
what a cancer diagnosis means while rejecting the usual platitudes about fighting cancer or
the notion that suffering somehow makes you stronger. Hitchens is brutally honest about

(46:10):
what cancer inevitably takes away from you, even those who survive longer than he did,
unfortunately. This isn't a comfortable read nor conventionally uplifting it. I first read
it a decade ago after my own cancer diagnosis and revisiting it recently I discovered new
insights that, while not exactly comforting, strengthened my resolve to continue working

(46:33):
against cancer for as long as I can.Well, thank you both for a fantastic
conversation. Really enjoyed it and really appreciate you being here.

JULI KLEMM (46:43):
Thanks, it was a lot of fun, Oliver.

Thanks, Oliver.

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. You can reach us at NCIICC@nih.gov.
Inside Cancer Careers is a collaboration between NCI’s Office of Communications and Public Liaison

(47:14):
and the Center for Cancer Training. It is produced by Angela Jones, Astrid Masfar, and Maria Moten.
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, you can email us at NCIinfo@nih.gov or

(47:36):
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.

All Episodes

Episode Transcript

Popular Podcasts

Crime Junkie

24/7 News: The Latest

Stuff You Should Know

.css-15opob5{left:0;position:absolute;top:0.8rem;} All Episodes

.css-14f5ked{margin:0;word-break:break-word;display:-webkit-box;-webkit-box-orient:vertical;box-orient:vertical;-webkit-line-clamp:2;overflow:hidden;}Revolutionizing Cancer Research: Navigating the AI Landscape

Episode Transcript

Popular Podcasts

.css-r6mb8g{margin:0;word-break:break-word;display:-webkit-box;-webkit-box-orient:vertical;box-orient:vertical;-webkit-line-clamp:1;overflow:hidden;}Crime Junkie

24/7 News: The Latest

Stuff You Should Know

All Episodes

Revolutionizing Cancer Research: Navigating the AI Landscape

Crime Junkie