Kimma: flexible linear mixed effects modeling with kinship covariance for RNA-seq data

Area of Science:

• Bioinformatics
• Genomics
• Statistical Genetics

Background:

• Identifying differentially expressed genes (DEGs) is crucial in transcriptomic research.
• Existing tools lack support for covariance matrices in DEG modeling, limiting analytical flexibility.

Purpose of the Study:

• Introduce kimma, an open-source R package for flexible linear mixed effects modeling.
• Enable DEG analysis incorporating covariates, weights, random effects, and covariance matrices.

Main Methods:

• Developed kimma, an R package implementing linear mixed effects models.
• Evaluated kimma's performance against established DEG tools using simulated datasets.
• Incorporated genetic kinship covariance matrices to assess their impact on DEG detection.

Main Results:

• kimma demonstrated comparable specificity, sensitivity, and computational time to limma and dream.
• kimma uniquely supports covariance matrices and fit metrics like AIC.
• Analysis of a related cohort revealed kinship's significant impact on model fit and DEG identification.

Conclusions:

• kimma offers a powerful and flexible alternative for DEG analysis, especially in related cohorts.
• The package equals or surpasses existing DEG pipelines in key performance metrics.
• kimma enhances the ability to model complex genetic architectures in transcriptomic studies.