SCassist: An AI Based Workflow Assistant for Single-Cell Analysis.

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Single-cell RNA sequencing (scRNA-seq) analysis is a complex, multi-step process demanding significant bioinformatics expertise and time.
• Existing workflows often present challenges for researchers, limiting accessibility and efficiency in biological data interpretation.

Purpose of the Study:

• To develop an R package, SCassist, that integrates large language models (LLMs) to streamline and enhance scRNA-seq data analysis.
• To provide researchers with intelligent, LLM-driven guidance for critical analysis steps and interpretation of results.

Main Methods:

• Developed SCassist, an R package utilizing LLMs (Google's Gemini, OpenAI's GPT, Meta's Llama3) for scRNA-seq analysis.
• Integrated LLM-powered recommendations for data filtering, normalization, and clustering parameters.
• Implemented LLM-guided interpretation of variable features, principal components, cell type annotation, and enrichment analysis.

Main Results:

• SCassist provides automated, data-driven recommendations for optimizing scRNA-seq analysis parameters.
• The package offers insightful, LLM-generated interpretations of complex genomic data, including feature significance and cell population identification.
• Demonstrated enhanced accessibility of sophisticated scRNA-seq analysis for researchers across various experience levels.

Conclusions:

• SCassist significantly reduces the complexity and time required for scRNA-seq data analysis.
• Leveraging LLMs in bioinformatics tools like SCassist democratizes advanced genomic data interpretation.
• The R package empowers researchers to conduct more robust and accessible scRNA-seq studies.