CS alumnus Alex Lu is a member of the Dean’s Advisory Council and a researcher at Microsoft New England. (Photo: Nick Iwanyshyn)

Computer science alumnus Alex Lu is working at the leading edge of the deep learning revolution, applying novel machine learning methods to generate hypotheses in biology.

Lu works as a senior researcher on the team, known affectionately as NERDs. Part of its core mandate is developing AI technology and understanding how technology impacts our lives.

“My research concerns how we can use artificial intelligence to accelerate discovery in the biological sciences,” says Lu, who earned his master’s degree in 2017 and his doctoral degree in 2021, both in computer science at U of T.

Lu’s research develops deep learning methods to help generate biological hypotheses. (Photo courtesy of Alex Lu)

Lu’s main expertise is microscopy images, but he works across a range of subjects including protein sequences, single-cell RNA, medical images and, most recently, the social sciences.

“Imagine replacing the manual effort of looking under a microscope with automated technologies that can easily clear hundreds of thousands, if not millions of images in a day,” Lu says.

The next question, and the one driving , is how that technology can be leveraged to formulate fascinating hypotheses and discover captivating biology without having to look at every data point.

On top of his work at Microsoft New England, Lu volunteers with U of T’s Faculty of Arts & Science as a member of the Dean’s Advisory Council, to collaborate with and draw on the remarkable breadth and depth of the A&S community and help the faculty fulfill its strategic priorities.

“It’s been an absolutely terrific opportunity,” Lu says.

As part of a highly interdisciplinary committee reflecting the composition of A&S itself, scientific experts like Lu meet with entrepreneurs, academics and thought leaders, sharing their diverse experiences as advisors to on a range of topics.

“And it’s not just expertise,” Lu says. “It’s the ability to provide a perspective from different vantages and that creates more innovative solutions than from a community formed of a more homogenous population.”

It’s a big responsibility, but Lu is up to the challenge. In his research, Lu has shown no hesitation to plunge into subject areas he may be less familiar with. Part of the appeal of his methods-based approach to science is the wide number of applications.

“It’s inspiring that he has the courage to work on something he’s never worked on before,” says , Lu’s doctoral supervisor from his time at U of T. “A lot of academics are afraid to do something like that.”

Moses is a computational biologist in the who is cross-appointed to the ������Ƶ of Computer Science. His work centres on understanding genomes, proteins and the molecular biology world. When Lu was pursuing his PhD, he had already developed an interest in microscope images, which made Moses’ lab a natural fit.

We are in the midst of a “deep learning” revolution, Moses says. His lab had long used traditional machine learning techniques but, during his PhD, Lu helped push the boundaries of what is possible in AI-driven microscopy by using deep learning models.

“Suddenly, we could do things with deep learning that had been either impossible or incredibly difficult with the machine learning tools we had before,” Moses says. “That was part of the reason Alex could really make revolutionary progress during his PhD.”

To illustrate what Lu accomplished, Moses raised the example of the human proteome — the entire set of proteins humans can express, which numbers around 20,000. Moses’ group examines microscope photos of the proteins in different cell types to learn how they behave. Using Lu’s deep learning models, which can handle far more complicated sets of parameters than a standard machine learning model, the team was able to easily pick out individual interesting proteins and create a comprehensive map of how they behave in cells.

“Alex helped us make a lot of progress because we could now analyze all 20,000 proteins at the same time,” Moses says. “It allows us to organize these very complicated things in a way we really couldn't do before.”

Lu has seamlessly transitioned his research at U of T into his work at Microsoft. Just as Moses empowered Lu to pursue a wide range of subjects during his PhD, so has Microsoft. The freedom to self-direct quickly became one Lu’s favourite parts of his work.

“My interest has always been in data, and in working with different kinds of data, not necessarily in answering a particular qualitative question,” Lu says.

Lu’s research has led him to some big insights into the future of machine learning, and he predicts the way scientists currently train models will soon change.

“We need more tailored approaches instead of just trying to brute-force as much data as you can with a one-size-fits-all method — and I'm hoping that leads to more thoughtful methods,” Lu says.

Adds Moses: “For me as a computational biologist, it's hard to overstate the excitement and the possibilities AI and deep learning are bringing to our research. Alex is really at the forefront of that revolution.”

— Original story by Coby Zucker for