Transcriptome-based cell type assignment for kidney cell culture models.

Area of Science:

• Nephrology
• Genomics
• Bioinformatics

Background:

• Kidney cell lines are crucial for studying kidney physiology and disease.
• However, gene expression in cell lines can diverge from primary cells due to various factors.
• Accurate validation of cell line identity is essential for reliable *in vitro* research.

Purpose of the Study:

• To develop and validate a transcriptome-based method for assessing kidney cell line identity.
• To compare the effectiveness of different statistical and machine learning approaches for matching cell line data to reference transcriptomes.
• To provide accessible tools for researchers to evaluate kidney cell model suitability.

Main Methods:

• Generated reference transcriptomic profiles from human and murine kidney single-cell RNA-seq (scRNA-seq) datasets.
• Matched bulk RNA-seq data from kidney tissues and cell lines to reference profiles using Spearman correlation, Euclidean distance, Poisson distance, and machine learning classifiers (Random Forest, XGBoost, TabPFN).
• Validated matching accuracy using global gene expression, kidney marker genes, and variable genes through a three-step strategy.

Main Results:

• Gene expression rank-based Spearman correlation and the TabPFN model demonstrated high accuracy and specificity, especially with curated kidney marker gene lists.
• These methods successfully identified microdissected kidney tubule segments and were robust against non-kidney controls.
• Specific kidney cell lines showed varying degrees of identity preservation; OK cells maintained proximal tubule identity under shear stress, while others were inconsistent. mIMCD-3 cells showed stable collecting duct identity.

Conclusions:

• A robust transcriptome-based approach using Spearman correlation or TabPFN can accurately assess kidney cell line identity.
• The developed tools, CellMatchR (web-based) and TabPFN scripts, facilitate informed selection and interpretation of kidney cell models.
• This methodology enhances the reliability of *in vitro* findings in kidney research.