Gene-Chronos: parameter-efficient developmental time inference using a pretrained single-cell foundation model.

Gene-Chronos accurately infers developmental time across diverse species, outperforming existing methods with a 20% improvement in robustness and generalization.

• Why it matters: Understanding continuous developmental timing from static single-cell data is crucial for unraveling complex biological processes, yet current methods lack robustness and interpretability across heterogeneous datasets.
• What they did: The approach employs a parameter-efficient framework built on a pretrained Geneformer model, incorporating learnable temporal prompts and contrastive learning to extract time-related signals from transcriptomic data.
• The result: Gene-Chronos achieves superior performance on multiple benchmarks, generalizes to unseen samples, and reveals key genes linked to developmental progression, enabling deeper insights into temporal gene expression dynamics.