LLM-scCurator: Data-centric feature distillation for zero-shot cell-type annotation
Furudate, K.; Takahashi, K.
Show abstract
Zero-shot cell-type annotation with language models degrades when marker lists are dominated by biological-noise programs (e.g., ribosomal/cell-cycle/stress). We present LLM-scCurator, a data-centric, backend-agnostic framework using pre-prompt noise masking and Gini-informed distillation to recover identity markers. Across benchmarks, LLM-scCurator improves ontology-aware hierarchical accuracy from 78.4% to 86.1% (52 clusters) over naive prompting, approaching supervised/reference-transfer performance without training data. LLM-scCurator extends to spatial transcriptomics (Visium, Xenium) for reference-free discovery of fine-grained niches.
Matching journals
The top 3 journals account for 50% of the predicted probability mass.
Similar papers in this journal
Similar papers in this journal
- High-throughput single-cell chromatin accessibility CRISPR screens enable unbiased identification of regulatory networks in cancer 98%
- GraphVelo allows for accurate inference of multimodal velocities and molecular mechanisms for single cells 97%
- Achieving single nucleotide sensitivity in direct hybridization genome imaging 97%
Similar papers in this journal
- Active repression of cell fate plasticity by PROX1 safeguards hepatocyte identity and prevents liver tumourigenesis 97%
- A global atlas of genetic associations of 220 deep phenotypes 96%
- Orphan CpG islands boost the regulatory activity of poisedenhancers and dictate the responsiveness of their target genes 96%
"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.