Emma Flynn, CPCB student
CPCB student Emma Flynn presents research at the Machine Learning in Drug Discovery Symposium

PhD candidate Emma Flynn recently took to the stage at the Machine Learning in Drug Discovery Symposium in Cambridge, Massachusetts, where she delivered a lightning talk. 

This symposium, which occurred on Nov. 1, highlighted progress in the application of machine learning techniques to drug discovery. Researchers working in machine learning shared ideas over a series of talks, poster sessions and networking opportunities. 

Flynn, who is a second-year student in the Joint Carnegie Mellon-University of Pittsburgh PhD Program in Computational Biology (CPCB), presented “PharmacoForge: Pharmacophore Generation with Diffusion Models.” This was a joint project with members of Associate Professor David Koes’ lab, including Riya Shah, Rishal Aggarwal and Ian Dunn. 

With this project, Flynn and the Koes lab are working on creating a diffusion model to generate new pharmacophores and find molecules that can match to proteins. Pharmacophores are used in drug discovery to identify potential lead compounds, and they can be useful when there is limited information known about a target. This generative modeling process has the potential to identify new drug leads. 

Emma Flynn stands beside a poster of her PharmacoForge project.

This was Flynn’s first research conference, and it proved to be a professional milestone for her. “The event was tailored to my interests, and it was exciting to see trends in the field,” she said. 

She also appreciated the chance to connect with other students and professionals in the machine learning community, many of whom stopped by her poster session after the talk to share feedback. 

“It gave me a lot of confidence because I was initially stressed for the presentation, but everything went well,” Flynn said. “I have a lot of new ideas for the future.” 

Looking ahead, Flynn and her collaborators are refining PharmacoForge. They aim to transition from a diffusion model to a flow-matching model, enhancing the tool’s accuracy. The Koes lab is also planning on submitting this research for journal publication, an exciting step for Flynn in sharing her findings with the wider scientific community. 

She is looking forward to more opportunities to share her work with others, and she leaves students interested in presenting at conferences with some advice: 

“Make sure you really know your presentation, because then you can just let the practice take over if you’re nervous.”