kimma

kimma models differential gene expression in transcriptomic (RNA-seq) datasets using linear mixed-effects models that incorporate covariance matrices (e.g., genetic kinship) to account for covariates, weights, and random effects.

Key Features:

• Flexible linear mixed-effects modeling: Integrates covariates, weights, and random effects into per-gene differential expression models.
• Covariance matrix support: Accepts covariance matrices, including genetic kinship, to account for relatedness and other structured covariance in models.
• Fit metrics (AIC): Provides model fit evaluation using the Akaike Information Criterion for model comparison and selection.
• Simulated dataset benchmarking: Demonstrated specificity and sensitivity comparable to limma unpaired and dream paired on simulated datasets while exhibiting similar computational efficiency.
• Kinship effect detection: Reveals the influence of genetic relatedness on model fit and differential expression detection in cohorts with familial or population structure.

Scientific Applications:

• Differential expression analysis: Identification of differentially expressed genes in transcriptomic/RNA-seq datasets while accounting for complex covariance structures.
• Family-based and population studies: Analysis of cohorts with genetic relatedness or population structure by incorporating kinship covariance.
• Genetics and evolutionary biology: Investigation of gene expression variation influenced by relatedness in genetic and evolutionary studies.
• Personalized medicine: Evaluation of expression differences in contexts where individual relatedness or structured covariance may affect inference.

Methodology:

Per-gene linear mixed-effects modeling with inclusion of covariance matrices (e.g., genetic kinship), fixed effects (covariates), random effects, and weights, and model fit assessed via Akaike Information Criterion (AIC).