New autoregressive transformer model for single-cell gene expression generation

Researchers have developed a novel autoregressive transformer model designed for generating single-cell gene expression vectors. This model, which incorporates a learned quantized variational auto-encoder tokenizer, is trained using a cross-entropy loss. The study evaluates the biological fidelity and scaling behavior of the model, identifying a compute-optimal frontier for single-cell foundation models. The pretrained model is also discussed as a potential tool for finetuning for perturbation response prediction. AI

IMPACT Introduces a new foundation model for single-cell data, potentially advancing biological research and drug discovery.

RANK_REASON Academic paper detailing a new model architecture and its scaling properties. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.AI →

• alphaXiv
• arXiv
• autoregressive transformer
• CatalyzeX
• DagsHub
• Gotit.pub
• Hugging Face
• IArxiv
• machine learning
• ScienceCast
• Single-Cell Generation
• variational auto-encoder

• paper
• model release

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →