Most tissue samples are composed of different cell types. Cell type composition can have large impact on RNA-seq or DNA methylation data collected from tissue samples. Ignoring cell type composition variation can lead to wrong conclusions. I will present two related topics to address cell type compositions. One is the estimation of cell type proportions using RNA-seq or DNA methylation data, including a recent work where we use RNA isoform expression to estimate cell type proportions. The other topic is cell type-specific differential expression analysis using bulk RNA-seq data. We have developed a method named CARseq that employs a negative binomial distribution that appropriately models the count data from RNA-seq experiments. CARseq performs cell type-specific differential expression analysis in a likelihood framework, which accounts for the uncertainty to estimate cell type-specific expression.