Learn with DLAG: A deep dive into deep learning

From the Deep Learning Affinity Group (DLAG)

Chatbots, and artificial intelligence more generally, have been dominating the headlines in recent months.

Just as a bench scientist learns through protocols and experience to execute a complex experiment, researchers can “teach” a machine to perform computational tasks using trained algorithms. This method, and the field at large, is called machine learning.

Deep learning, a subset of machine learning, utilizes neural networks to improve prediction accuracy of computational models. Neural networks, modeled after biological neural networks in animal brains, contain an interconnected group of artificial neurons and nodes. Neural networks used for deep learning often contain many layers of artificial neurons in their network architecture (hence the moniker “deep”). At Fred Hutchinson Cancer Center, a number of labs have turned to machine learning to solve a range of scientific questions, from developing new tools for cancer diagnostics to predicting inherited colorectal cancer risk.

The Deep Learning Affinity Group (DLAG) is a brand-new affinity group at Fred Hutch for researchers with an interest in learning about and applying deep-learning methods. DLAG hosts a monthly seminar series, and the only prerequisite to participate is an interest – not necessarily deep knowledge – in deep learning.

What is the Deep Learning Affinity Group (DLAG)?

DLAG is an affinity group born out of the shared interests of Fred Hutch faculty members Drs. Youyi FongJeff LeekEvan NewellWei Sun, and Michael Wu.

An email conversation among the five in March 2023 revealed that an affinity group at the Hutch dedicated to deep learning, modeled after similar groups including the Clinical Trials Affinity Group (CTAG) and Math Modeling Affinity Group, could serve an unmet need.

“We formed this group with the intent to help each other better understand and harness the power of deep learning, whose growth has been truly phenomenal,” said Dr. Youyi Fong, a professor in the Vaccine and Infectious Disease Division and the Public Health Sciences Division.

DLAG hosts a monthly seminar series featuring both Fred Hutch and external researchers from Seattle and beyond. The speakers cut across different departments and areas of scientific expertise, but share the common ground of utilizing deep learning in their research.  

“We hope this platform can attract more people, inside or outside the Hutch, with different backgrounds to work together on interesting deep learning related problems,” said Dr. Wei Sun, a professor in the Biostatistics Program and the Public Health Sciences Division.

“Everyone is welcome”

As an affinity group to encourage the exchange of ideas and learning, DLAG has no formal membership. Everyone is encouraged to join, including trainees and those with limited knowledge of deep learning.

“Everyone is welcome to join the seminar and discussions,” said Dr. Sun. “We also welcome suggestions on topics to discuss or potential speakers, or sharing your deep-learning papers that we can list on the external website.”

DLAG seminars take place monthly in person or via Zoom, typically at noon on the first Tuesday of the month. In the future, DLAG plans to organize workshops and symposiums to share ideas and encourage innovation, Dr. Fong said.

An AI-generated abstract painting of a woman in a lab coat typing on a typewriter in a science laboratory.
DALL-E’s interpretation of the author writing this article. The same deep-learning model architecture underlying AI programs like ChatGPT and DALL-E is often used for solving biological questions. DLAG is a brand-new affinity group at Fred Hutch for researchers interested in learning about and applying deep-learning methods. Image generated by the author using DALL-E 2

Monthly DLAG seminars

DLAG kicked off its seminar series in May with the inaugural speaker, Dr. William “Bill” Stafford Noble from the Department of Genome Sciences at the University of Washington. Dr. Noble discussed two applications of deep learning for mass-spectrometry proteomics, where a core problem is inferring the original protein sequence based on mass-spectrometry readouts.

The interdisciplinary nature of DLAG provides a space for fruitful discussions and opportunities for cross pollination of ideas. “Even though I don't typically encounter mass spec data, the way Bill solves the problem using deep learning gave me a lot of food for thought,” said Dr. Fong, who conducts research on infectious diseases as a bioinformatician.

DLAG, and by extension the speakers for the seminar series, includes scientists working on a diverse range of research topics. “This is a reflection of the fact that deep learning has been a driving force in innovations in many, many fields,” said Dr. Fong.

Next month’s seminar is Thursday, June 15 at noon via Zoom. Dr. Young Hwan Chang, an associate professor at the Oregon Health and Science University, will give a talk titled “Representation Learning and its Application on Multiplex Tissue Imaging Data.”

Recordings of past seminars are available on CenterNet.

Get involved

External website: https://research.fredhutch.org/dlag/en.html

CenterNet website: https://centernet.fredhutch.org/cn/u/dlag.html

To get updates on upcoming seminars, join the DLAG listserv.

Contact the organizers at dlag@fredhutch.org.

Noelle Noble, as well as Jill Anderson, executed administrative tasks involved in starting a new affinity group at the Hutch and were instrumental to getting DLAG off the ground.

Milly Jeffries and Jeremy Webb built the internal and external DLAG websites. 

Other Fred Hutch staff critical to launching and maintaining DLAG include Carol Wallace, Michael Torza, Vinthia Wirantana, Carlyn Fausto, and Melissa Alvendia; as well as the Public Health Sciences Division, and Vaccine and Infectious Disease Division leadership.