The inference of gene co-expressions from microarray and RNA-sequencing data has motivated many methodology developments for high-dimensional data and led to rich insights on biological processes and disease mechanisms. However, the bulk samples analyzed in most studies are a mixture of different cell types. As a result, the inferred co-expressions are confounded by varying cell type compositions across samples and only offer an aggregated view of gene regulations that may be distinct across different cell types. In this talk, we introduce two complementary statistical methods for inferring cell-type-specific co-expression networks based on two distinct sources of RNA sequencing (RNA-seq) data. First, to address the unique opportunity and challenge from the recently developed single cell RNA-seq (scRNA-seq) technology, we developed a method, named CS-CORE, that explicitly addresses the high sequencing depth variations and measurement errors present in single cell data for estimating and testing cell-type-specific co-expression. When applied to analyze multiple scRNA-seq datasets including those on Alzheimer’s disease, CS-CORE identified cell-type-specific co-expressions and differential co-expressions that were more reproducible and/or more enriched for relevant biological pathways than those inferred from the existing methods. Moreover, to leverage the rich collection of bulk RNA-seq data, we also developed CSNet, a flexible framework to estimate cell-type-specific gene co-expression networks from bulk sample data. We show that the proposed sparse least squares estimator is efficient to implement and enjoys good theoretical properties. When applied to analyze bulk RNA-seq data from Alzheimer’s disease, CSNet identified previously unknown cell-type-specific co-expressions among Alzheimer’s disease risk genes, suggesting cell-type-specific disease pathology for Alzheimer’s disease. This is joint work with Chang Su, Jingfei Zhang, Zichun Xu, Xinning Shan, and Biao Cai.