Cell-Type-Specific Proteogenomic Signal Diffusion for Integrating Multi-Omics Data Predicts Novel Schizophrenia Risk Genes

Abstract

SUMMARY Accumulation of diverse types of omics data on schizophrenia (SCZ) requires a systems approach to model the interplay between genome, transcriptome, and proteome. We introduce Markov affinity-based proteogenomic signal diffusion (MAPSD), a method to model intra-cellular protein trafficking paradigms and tissue-wise single-cell protein abundances. MAPSD integrates multi-omics data to amplify the signals at SCZ risk loci with small effect sizes, and reveal convergent disease-associated gene modules in the brain. We predicted a set of high-confidence SCZ risk loci followed by characterizing the subcellular localization of proteins encoded by candidate SCZ risk genes, and illustrated that most are enriched in neuronal cells in the cerebral cortex as well as Purkinje cells in the cerebellum. We demonstrated how the identified genes may be involved in neurodevelopment, how they may alter SCZ-related biological pathways, and how they facilitate drug repurposing. MAPSD is applicable in other polygenic diseases and can facilitate our understanding of disease mechanisms.

Tissue-wide cell-specific proteogenomic modeling reveals novel candidate risk genes in autism spectrum disorders

“…We used MAPSD 31 , a multi-omics data integration method, to identify additional candidate genes for ASD from existing knowledge. MAPSD first identifies omics information (genome, transcriptome, proteome) on known candidate genes for a specific disease from various sources; it then receives PPI networks, subcellular localization of proteins within cellular micro-domains, and protein abundances across >130 different combinations of tissues and cell-types as well as a repertoire of other omics data-types followed by diffusing the accumulating signal intensities of available genetic signatures through tissue/cell-type adjusted PPI networks, to uncover disease-relevant genetic drivers of the disease with small effect sizes which cannot be captured using available single-omics pipelines (Fig.…”

“…A good example of such diseases is schizophrenia (SCZ) in which high polygenicity of the disease has been extensively studied in the literature 32 . Our previous study on SCZ had revealed a shortlist of novel disease associated candidate risk genes with a similar enrichment patterns in specific cellular micro-domains in neuronal cells in human cerebral cortex 31 . Given the availability of valuable resources including distinct molecular data on ASD, this is a great opportunity to account for the complexities of ASD through a novel systems-level approach.…”

“…To address these questions, we have employed our develop method called MAPSD 31 , Markov Affinity-based Proteogenomic Signal Diffusion, for joint modeling of biochemical and biophysical properties of proteins with genomic and transcriptomic signatures of ASD aimed at identification of novel ASD candidate risk genes. Upon collecting multiple sources of omics data from the literature, MAPSD is able to generate tissue and cell-specific interactomes followed by diffusing the known biological signals throughout the protein-protein interaction networks (PPI) networks in order to predict unannotated susceptible genes.…”

A computational method for direct imputation of cell type-specific expression profiles and cellular compositions from bulk-tissue RNA-Seq in brain disorders

“…Recent advancement in single cell RNA-seq (scRNA-seq) technologies has made it clear how specific cell types affect the diseases mechanisms. Remarkable findings in autism spectrum disorders ( 8 ), schizophrenia ( 1 , 9 , 10 ), studying retinal tissue ( 11 ) and anatomy of human kidneys ( 12 ) all demonstrated how specific cell types are most relevant to the pathogenesis of different diseases.…”

Cell Type-Specific Annotation and Fine Mapping of Variants Associated With Brain Disorders

“…The majority of complex diseases are context-specific in that not every tissue is equally vulnerable to the genetic variation, while most of the available predictive measures do not take into account such information. For example, in the case of neuropsychiatric diseases such as schizophrenia (SCZ), genetic variants with transcriptional effects in the central nervous system may have small or no effects in other tissues, so tissue-specific information can facilitate the identification of variants that play causal roles in disease pathogenesis ( Skene et al, 2018 ; Mendizabal et al, 2019 ; Doostparast Torshizi et al, 2020 ). Moreover, genetic variations are known to play a central role in conferring susceptibilityof autism spectrum disorders (ASDs) and other neurodevelopmental disabilities ( Sanders et al, 2015 ; Doostparast Torshizi et al, 2018 ; Grove et al, 2019 ).…”