Single-cell interactomes of the human brain reveal cell-type specific convergence of brain disorders

The reference human interactome has been instrumental in the systems-level study of the molecular inner workings of the cell, providing a framework to analyze the network context of disease associated gene perturbations. However, reference organismal interactomes do not capture the tissue- and cell type-specific context in which proteins and modules preferentially act. Emerging single-cell profiling technologies, which survey the transcriptional cell-state distribution of complex tissues, could be used to infer the single-cell context of gene interactions. Here we introduce SCINET (Single-Cell Imputation and NETwork construction), a computational framework that reconstructs an ensemble of cell type-specific interactomes by integrating a global, context-independent reference interactome with a single-cell gene expression profile. SCINET addresses technical challenges of single-cell data by robustly imputing, transforming, and normalizing the initially noisy and sparse expression data. Subsequently, cell-level gene interaction probabilities and group-level gene interaction strengths are computed, resulting in cell type specific interactomes. We use SCINET to analyze the human cortex, reconstructing interactomes for the major cell types of the adult human brain. We identify network neighborhoods composed of topologically-specific genes that are central for cell-type influence but not for global interactome connectivity. We use the reconstructed interactomes to analyze the specificity and modularity of perturbations associated with neurodegenerative, neuropsychiatric, and neoplastic brain disorders; finding high variability across diseases, yet overall consistency in patterns of cell-type convergence for diseases of the same group. We infer for each disorder group disease gene networks with preferential cell-type specific activity that can aid the design and interpretation of cell-type resolution experiments. Finally, focusing on the pleiotropy of schizophrenia and bipolar disorder, we show how cell type specific interactomes enable the identification of disease genes with preferential influence on neuronal, glial, or glial-neuronal cells. The SCINET framework is applicable to any organism, cell-type/tissue, and reference network; it is freely available at https://github.com/shmohammadi86/SCINET.