Using Machine Learning To Revolutionize Drug Discovery
Xuezhu Cai, PhD’20, bioengineering, used her LEADERs fellowship with Merck to develop a new machine learning pipeline to ensure accurate data in drug discovery. She is now a senior scientist there.
Xuezhu Cai, PhD’20 Followed Her Passion for Advanced Machine Learning Into a LEADERs Fellowship with Merck. Now She’s a Senior Scientist There.
It used to take ten years to discover a new drug, but today’s game-changing sequencing methods, like CRISPR screening, allow scientists to identify pharmaceutical targets in half that time or even less. Thanks to machine learning, a large volume of medical images can be analyzed for not one but thousands of potential compounds (across the entire genome) all at once.
The question now is how do we ensure the data is accurate?
That’s why LEADERs alum Xuezhu Cai, PhD’20, bioengineering, now a senior scientist at Merck after completing her LEADERs fellowship with the multinational pharmaceutical leader, has been developing a new accelerated machine learning pipeline. She’s applying deep learning (a network that teaches computers to more accurately process data) to create and analyze clearer multi-dimensional images of cells that may be overlapping or clumping together, for example.
“The traditional way of machine learning is not enough to extract robust features of some organisms,” says Cai. “The old algorithms and statistical models lack data for us to train on or label, so there’s a need to continuously use new advanced models over the models that were developed a decade ago.”
Rewind five years and Cai found herself searching for a needle in a different haystack: her career path. As she decided how to enter the field as a doctor of philosophy, she learned about the Northeastern PhD Network’s LEADERs program from her advisor Craig Ferris, who had been mentoring her in the Center for Translational Neuroimaging. It had been up to Cai to find new connections between different parts of the human brain in neurodegenerative diseases like Alzheimer’s and Parkinson’s (tracking data with partner institutes between 173 brain images with the Brain Atlases project). This was a big job and Ferris was acutely aware of how promising the future was for his talented, young research assistant.
“It was a daunting task. She was marvelous at it,” says Ferris, adding that Cai’s work has since been published in over a dozen journals. “Xuezhu came to us from Yale’s biomedical engineering master’s program and she excelled at working with big numbers.”
At Ferris’ urging, Cai signed up for the LEADERs course, “Leading Self and Others,” 7600, and, from it, walked away with new core management and workplace development skills. She emerged as a natural-born leader and people noticed. Cai was placed in a LEADERs fellowship
at Merck, where she was later hired fresh out of graduation as a senior scientist, making her passion for advanced machine learning the focus of her career.
“That’s the beauty of the LEADERs program. It puts students in industry areas that they’re interested in, even though it may not be exactly what they’re doing in the laboratory,” says Ferris. “It gives them a whole new skillset and it also opens the door immediately for them, through a placement in industry, to be potentially hired afterwards.”
Now, at the bustling intersection of healthcare and predictive analytics (where continued massive job growth is expected), in research and development in Merck’s Image Analytics IT department, Cai is solving big data problems outside a university lab.
“Merck is very research intensive, so it feels like you’re doing a PhD program. It’s like an academic environment,” says Cai. “I had some exposure to confocal microscopy [advanced imaging] during my PhD research, so that opened the door to my LEADERs fellowship. Now I’m dealing with medical imaging from early discovery all the way to the clinical phase.”
She says that there are so many jobs out there for anyone interested in solving new biomedical engineering challenges by way of emerging technologies on this new path to drug discovery. That includes everything from continuous improvement of deep learning to scaling images so they can be transferred onto the cloud for broader, wider, and quicker scientific analyses.
“In high-content screening, cells don’t always act the same way from cell line to cell line, so there are a lot of problems to be solved,” she says. “There are a lot of places where people with this kind of skill can help provide value, as we migrate from traditional methods to deep learning.”
Upon completion of the “Leading Self and Others” course, LEADERs program staff and partner companies select fellows who align with specific industry needs. PhDs are supported through the fellowship application process and placed in a specialized role to solve a problem in industry. They go on to earn a LEADERs’ Experiential PhD leadership certificate, with guidance from an industry mentor and faculty advisor. The program is run by the PhD Network, which helps prepare students to enter the workforce as impactful researchers.
Source: PhD Network by Anna Fiorentino