lmFit

lmFit is a core function in the limma package (Bioconductor) used for fitting linear models to gene expression data across samples. For each gene, it estimates coefficients corresponding to the effects specified by a design matrix, such as treatment versus control, timepoints, or other covariates. The input is typically an expression matrix with genes as rows and samples as columns, together with a design matrix that encodes the experimental design. The function can also incorporate information to account for correlation between samples in paired or blocked experiments.

lmFit returns an object (commonly of class MArrayLM) containing estimated coefficients, standard errors, residuals, and fitted values for every gene. This object serves as the basis for downstream steps in the limma workflow, such as specifying contrasts with contrasts.fit or makeContrasts to define comparisons of interest, and applying empirical Bayes moderation via eBayes to obtain stabilized variance estimates and moderated statistics.

Common downstream steps include extracting results with topTable to obtain a ranked list of differential expression and using decideTests for rule-based decisions about significance. In RNA-seq analyses, limma is frequently used with the voom transformation, which converts counts to log-counts per million with a mean-variance trend, enabling the use of lmFit on non-microarray data.

lmFit supports complex experimental designs and can handle correlated samples through appropriate design specifications and correlation structures. It is a foundational tool in the limma differential expression workflow and is widely used in genomics research for identifying gene-level effects across multiple conditions or covariates.