CCB Seminar: Decoding Cellular Identities from Single-Cell Data
Speaker: Zoe Piran, The Hebrew University of Jerusalem
Title: Decoding cellular identities from single-cell data
Molecular profiling of a cell captures multiple overlapping signals arising from its internal state as well as possible external interventions. Focusing on a single signal (measured or recovered) restricts the ability to unveil the intricacies and diverse attributes of cellular states. To address this, we suggest computational frameworks to decode the rich layers of information encoded in single-cell genomics data. First, we introduce “SiFT” [1], a probabilistic computational framework that utilizes existing prior knowledge to filter known signals uncovering additional underlying biological attributes. To further decouple the different biological signals, we developed “biolord” [2], a deep-learning method for disentangling single-cell genomics data into the multiple facets of cellular identity. The obtained representation allows the generation of experimentally inaccessible cell states, such as accurate prediction of response to unseen drugs. As we demonstrate over a diversity of tasks, these methods allow us to uncover and study different facets of cellular identities. For example, when applied to COVID-19 data “SiFT” exposes disease-related dynamics, and “biolord” exposes infection-related signals in an atlas of Plasmodium infection.
- Z. Piran & M. Nitzan, “SiFT: uncovering hidden biological processes by probabilistic filtering of single-cell data.” Nature Communications 15.1 (2024): 760. https://doi.org/10.1038/s41467-024-44757-7.
- Z. Piran, N. Cohen, Y. Hoshen & M. Nitzan, “Disentanglement of single-cell data with biolord.” Nature Biotechnology (2024): 1-6. https://doi.org/10.1038/s41587-023-02079-x.