BMI Professor Christina Kendziorski and her lab report CASSIA, a new multi-agent large language model for automated and interpretable cell annotation in Nature Communications.
Large-language models have potential to improve cell annotation, but face challenges, including hyperconfidence, hallucinations, and lack of reasoning.
Dr. Kendziorski’s team developed CASSIA to address these challenges, providing for automated, accurate, and interpretable cell annotation of single-cell RNA-Seq data.
CASSIA improves annotation accuracy in both benchmark data sets and complex and rare cell populations, while also providing users with reasoning and quality assessment to ensure interpretability, guard against hallucinations, and calibrate confidence.
Congratulations to Christina and her team!