Emerging frontiers for AI in single-cell genomics.

AI models are increasingly being exploited to interpret large-scale single-cell genomics datasets, especially as atlas efforts and multiomic technologies expand.

• Why it matters: Understanding the vast and growing single-cell data resources is crucial for advancing biological insights, but early AI embedding models did not gain widespread adoption among experimental biologists.
• What they did: The authors review AI embedding and foundation models for scRNA-seq and scATAC-seq, machine-learning approaches for single-cell multiome data, and AI strategies for learning "virtual cell" models from Perturb-seq, highlighting current limitations and future prospects.
• The result: This work underscores the potential of AI to enhance interpretation of single-cell data and guides future developments to overcome existing model limitations in the field.