Description:

OCTOBER MEETING: Thursday, October 30, 2025 at 5:30pm (US Central)

LOCATION: 
In person: Chambers Hall, Lower Level
600 Foster Steet, Evanston Campus

AGENDA:
5:30pm - Meet and greet with refreshments
6:00pm - Talk with Buduka Ogonor, Motter Group, Department of Physics and Astronomy, Center for Network Dynamics

TALK TITLE:
Finding gene sets underlying complex phenotypes via generative modeling

ABSTRACT:
Most human traits are complex, that is, they emerge from the interactions among multiple genes. The sheer number of possible gene set-to-phenotype mappings makes it challenging to identify the gene sets underlying complex traits using statistical approaches like genome and transcriptome-wide association studies. Furthermore, these existing approaches assume that molecular-level changes are independent, which is at odds with the existence of intracellular networks that govern cell behavior.

Here, we present an approach that identifies gene set-to-phenotype relationships that leverages generative modeling trained on publicly available transcriptional data. We use a generative model—dubbed TWAVE for Transcription-Wide Variational Auto-Encoder—to emulate diseased and healthy transcriptional states. Then, we use existing transcriptional measurements of responses to turning genes on and off as inputs to an optimization framework, which identifies the gene perturbations that minimize the transcriptional difference between the diseased and healthy states. Using nine disease traits as examples, we show that the approach identifies causal genes that cannot be detected by the primary existing techniques. We suggest that the approach be used to design tailored experiments to identify multi-genic targets to address complex diseases.

DATA SCIENCE NIGHTS are monthly meetings featuring presentations and discussions about data-driven science and complex systems, organized by Northwestern University graduate students and scholars. Students and researchers of all levels are welcome! For more information: http://bit.ly/nico-dsn