Creating tools to generate data insights

#WhyIScience Q&A: A computer scientist on the future of machine learning in biology

kzusi@broadinstitute.org — Thu, 14 Jan 2021 20:15:11 +0000

#WhyIScience Q&A: A computer scientist on the future of machine learning in biology

By Karen Zusi-Tran January 14, 2021

#WhyIScience Q&A: A computer scientist on the future of machine learning in biology

Schmidt Fellow Juan Caicedo discusses the benefits of interdisciplinary research, how to mentor a team during a pandemic, and how he’s helping to fight racism.

By Karen Zusi-Tran

January 14, 2021

Credit: Allison Dougherty, ӳ��ý Communications

Juan Caicedo develops machine learning tools that can automatically identify, extract, and analyze complex patterns in vast amounts of biological data.

Crafting a research agenda, learning how to mentor others, building leadership skills — these were the traditional challenges of starting a new lab that Juan Caicedo was focused on in late 2019. He never expected a pandemic to erupt just as he was getting his team off the ground.

As a Schmidt Fellow at the ӳ��ý, Caicedo develops machine learning tools that can automatically identify, extract, and analyze complex patterns in vast amounts of biological data. Sometimes this means analyzing millions of pictures of cells; other times it might be chemical structures, or gene expression data from single cells. No matter the data type, for Caicedo, it’s always about creating new technology to drive biological research and discovery.

Caicedo grew up in Colombia, just outside the capital city of Bogotá, and knew by the end of high school that he wanted to be a computer scientist. At the National University of Colombia, he was introduced to the field of artificial intelligence and machine learning. After completing his master’s and PhD in Colombia, he joined the University of Illinois as a postdoctoral researcher, and from there came to the lab of Anne Carpenter, senior director of the Imaging Platform at the ӳ��ý. In summer of 2019, he became a Schmidt Fellow, granted the opportunity and support to start his own lab at the institute.

We spoke with Caicedo about what excites him in machine learning, some of the challenges of putting together and leading a new team remotely, and his passion for diversity and equity in the research community.

Q: You started off in computer science. How did you find yourself gravitating towards biological questions?

During my master’s, I worked with medical doctors and pathologists who wanted to automate the analysis of diagnostic images. This was the first time I was invited to enter the biological labs. And I saw a lot of potential for using computer science and artificial intelligence to help them achieve their goals.

After that experience, I wanted the opportunity to work with people who are doing serious biological projects, but who are also open to automating, or want to improve the techniques and technologies that are being used. I also wanted to find something that was meaningful and useful to society, and I think biomedical applications have that component.

Q: What are you excited about in the future of machine learning in biology?

Great science is beautiful and inspiring in itself, and any way in which I could contribute to making science even more beautiful and useful motivates me. I'm thinking about tools that can support biological discovery: Is there an opportunity to make a discovery using machine learning in a way that we cannot achieve at the moment? Just like microscopes are a ubiquitous tool, I hope that someday, biology also has machine learning tools to compose information, combine data, and understand patterns.

Images give us a lot of information about the morphological state of cells, which reveals their structural organization. We know that structure determines function — thus, imaging is a powerful tool to understand cellular processes and an exciting opportunity.

One of the major challenges of studying imaging data is the association of pixel values with meaningful biological patterns. Think of recognizing when faces are happy or sad in pictures of people: modern computer vision systems can learn to do just that. The problem with images of cells is that sometimes we don’t know what a “smiling” cell looks like, or how many other expressions they may have. I think machine learning can help us with these challenges to measure the morphological expression of cells, and to support biological discovery.

Q: Several months into your fellowship, in spring 2020, ӳ��ý wound down most of its lab-based research due to COVID-19. That work has since resumed, but your team is still working remotely along with other computational teams. What has it been like to continue supporting your new lab under these conditions?

At the beginning of the year, it was an exciting time to start creating my lab and my team, and I had the very traditional type of lab in mind: We have meetings, and we are next to each other, and we can talk anytime. But then, of course, all the challenges came. At some point I realized that this vision of our lab that I had at the beginning was definitely not going to happen, but that doesn't mean we cannot have a team.

At some point I realized that this vision of our lab that I had at the beginning was definitely not going to happen, but that doesn't mean we cannot have a team.

We have to change the dynamics in many ways and open the channels of communication — like making everything written instead of relying on informal communications with people that you have by your side. This type of remote setting requires you to be very descriptive when you write any documentation.

We have also tried many different things to keep people engaged. I want people in my team to feel that they are connected to each other — like after a meeting, we randomly pair them to simulate that they are bumping into each other. We try to make everyone feel that there are real people behind this team. It's not just the text that we read or the code that we write.

Q: What are some of the hurdles that you've overcome in your career?

One challenge that I have faced, especially at the ӳ��ý, has been the challenge of imposter syndrome. There are so many extraordinary people here, sometimes you ask yourself if you’re going to produce the same amount or quality of results. And of course, in order to overcome those types of feelings and challenges, when the pressure builds up and expectations are high, it’s important to have the right people by your side, such as mentors and friends, people who have your back and encourage you and cheer you up. You never can do this alone.

Moving here to the US was exciting, but it was also challenging. And I think when you leave behind your family, your friends, you are putting yourself in a vulnerable place, in which anything can become more difficult. I had to learn English and I had to learn the culture and the traditions here, and that can be challenging. Pursuing a career in science requires a lot of effort and sometimes sacrifices.

But even though I left my family and friends in Colombia, I found all the people here who have always been willing to help me and support me — that's what keeps me going forward. My wonderful colleagues and wonderful friends here inspire me and remind me that nothing is impossible, especially if we do it together.

Q: What advice would you give to trainees or other young scientists?

If you are a biologist, you don't have to be afraid of machine learning or computer science or programming, and if you are a computer scientist, you don't have to be afraid of biology. We can connect with people who can help us understand and make sense of new fields. And, don't worry if you don't know everything, because nobody knows everything. The magic is in collaborating. Be open to learning and always be curious and look for ways to grow.

Don't worry if you don't know everything, because nobody knows everything. The magic is in collaborating.

For students who might come from similar backgrounds as mine, I would also especially emphasize how important it is to always believe in yourself. One of the reasons we see the United States as an attractive place to come is because there is this idea that as long as we give our best, we're going to have opportunities to grow here. But that doesn't mean it's going to come without issues or difficulties. And one of those is definitely the problem of discrimination. Personally, I think if you're ever the target of aggressions or discrimination, don't take it personally or as something final that is defining you. And, I'm not sure how to say this, but — be forgiving as well. Sometimes it's more lack of knowledge rather than evil.

Q: You are also participating in a ӳ��ý program to champion inclusion, diversity, equity, and allyship across the institute. What does this program mean to you?

Coming from an international background, I think that has helped me to understand the diversity of communities, from different perspectives. And even though academia is very open to diversity, there is still a gap to fill. But as I said, I like to see the problem as more of a lack of knowledge. In the same way as we don't know how to solve certain biological problems or some science problems, there are many people who don't know how to approach issues of racism. And I think having conversations and being open-minded is the only way to explore new possibilities and break old habits.

I feel strongly about defending people who don't have the same rights or opportunities. I come from a background that was not necessarily privileged. Public universities and public education gave me these opportunities to access a completely new world of professional development. I try to give back and speak in support of those who don't have the same opportunities.

And finally, as someone who comes from a minority group, I also have faced many of those challenges myself, being discriminated against sometimes or being the target of microaggressions. And none of that is openly discussed. We try to just brush that off, and try to keep going and move forward. But it's wrong to discriminate, it's wrong to make fun of people, it's wrong to attack people. It always starts with very minor actions, but it can escalate all the way to the point in which your life is threatened, and that's not right. And I think that type of attitude, that type of mentality, should change.

I see this program as an opportunity to open those discussions, to learn from each other, and to make the ӳ��ý an anti-racist organization, which I think is an amazing goal to have. I feel very, very proud of the ӳ��ý community for engaging in this type of conversation and being willing to take steps and gather as a community to discuss what could be done and experiment with new things. This goal is very high and very ambitious. It’s not easy. And it's a conversation that is going to be very difficult to have, and it's going to make us all uncomfortable. But it's important and it's a good cause to fight for, because it's the right thing to do.

Tags:

Creating tools to generate data insights

Image-Based Profiling

ӳ��ý and Verily partner with Microsoft to accelerate the next generation of the Terra platform for health and life science research

kzusi@broadinstitute.org — Mon, 11 Jan 2021 14:06:30 +0000