Generation of single cell multi-omics computational methods for biological insights and novel deconvolution techniques for bulk omics data

学生战略优先领域：生物能源和工业生物技术关键词：机器学习；单细胞生物学；去卷积；刺激；皮肤近年来，已经产生了许多大量的RNA-Seq和表观遗传学数据，但我们缺乏对细胞类型和组织位置信息的分辨率来得出强有力的结论。单细胞转录物测序(scRNA-Seq)的进展迅速，正在彻底改变人们理解细胞异质性和细胞间通信的能力。然而，到目前为止，还没有有效的方法来结合scRNA-Seq、单细胞开放染色质捕获(ScATAC)和空间转录(基因表达的组织学定位)，并使用它们来更好地理解现有的大量组学数据。例如，特定细胞中染色质开放程度的增加应该与大量数据中发现的同一基因组区域甲基化和表达的变化有关。目标：1)生成新的计算模型，以整合基因表达(scRNA-Seq)、表观遗传(ScATACseq)和信使核糖核酸(MRNAs)组织位置数据(空间转录)，以生成细胞特定的组织定位特征和多组网络。3)使用1中的模型和2中的签名来分解现有的RNAseq和甲基化数据，以将差异签名映射到细胞签名和多组网络上。4)在皮肤的单个细胞水平上识别刺激签名及其与皮肤状况的关系。

Studentship strategic priority area:Bioenergy and Industrial BiotechnologyKeywords: Machine learning; Single cell biology; deconvolution; irritation; SkinIn recent years, many bulk RNA-Seq and epigenetic data has been generated, but we lack the resolution of cell type and tissue positional information to draw strong conclusions.Advances in single-cell transcriptomics sequencing (scRNA-Seq) have been rapid and is revolutionizing the ability to understand, for example, cellular heterogeneity and cell-cell communication . To date, however, there are no efficient methods to combine scRNA-Seq, single-cell open chromatin capture (scATAC) and spatial transcriptomics (histological location of gene expression) and use them to better understand existing bulk 'omics data. For instance, an increase in chromatin opening in specific cells should link to changes in methylation and expression in the same genomic region found in bulk data. We propose to generate a multi-omics single-cell dataset from skin "cutaneous irritation", as an example biological domain, to build and test models for integration and interpretation of these multi-omic single-cell technologies that can be used to better interpret existing bulk omics data.Objectives:1) Generate novel computational models to integrate gene expression (scRNA-Seq), epigenetic (scATACseq) and mRNA tissue location data (spatial transcriptomics) to generate cell-specific tissue-localised signatures and multi-omic networks. 2) In-silico 'bulkify' cell signatures (2a) and validate with the actual bulk RNAseq data (2b).3) Deconvolute existing RNAseq and methylation data using models from 1 and signatures from 2, to map differential signatures onto cell signatures and multi-omic networks.4) Identify irritation signatures at a single cell level in skin and its relationship with skin conditions.