A web-based platform for robust single-cell analysis, bulk data deconvolution and system-level analysis

项目总结 以及测量数百万个单个细胞、单细胞 技术也带来了将这些数据转换为更好地理解 潜在的生物现象。用于单元格数据分析的现有计算方法和软件 在可伸缩性、准确性、可用性和解释能力方面存在严重限制。的主要目标是 该项目旨在开创一个新的单细胞数据分析平台，该平台能够：i)准确 鉴定复杂组织中的细胞类型及其组成，II)推断细胞发育阶段和假性 时间轨迹，以及iii)在表型中识别特定细胞类型的路径和假定的机制 比较一下。拟议的平台还将能够对批量表达数据进行去卷积以识别细胞类型 每个散装样品的组成。拟议工作的意义在于它有可能提供新的 用于单细胞数据分析的方法远远超过当前最先进技术的性能。 准确的反卷积还将使研究人员能够从巨大的数据仓库中提取更多信息 现有批量数据，包括GDC/TCGA、NCBI SRA、GEO和ArrayExpress，它们当前包含 来自批量实验的数据，总共耗资超过10亿美元。推动这项工作的假设是 批量数据的单单元数据分析和蜂窝去卷积可大大受益于：i)系统级 掌握细胞发展的关键特征的知识，以及ii)在 经过验证的单元格类型和参考单元格地图集中提供的单元格数据集。事实上，我们初步的 工作表明，单细胞数据分析和细胞去卷积可以达到卓越的精度 如果我们正确利用参考单细胞数据集和通路知识，大约有90%-100%。这个 建议的平台将通过将其功能与最先进的软件进行比较来进行广泛验证 在单细胞数据分析中(细胞类型识别、发育状态和时间轨迹推断， 系统级分析)和批量表达数据的细胞去卷积。这将使用两个663来完成 代表279种细胞类型和116个人体器官部分的数据集(包括批量数据、单细胞数据和 配对的细胞流式细胞术)。路径分析和机制推理能力将进一步提高 使用真实的基因敲除数据集(其中表型的真正原因已知)进行验证。这家公司， Advaita，拥有强大的IP组合，经验丰富的团队，以及在该领域经过验证的跟踪记录，已开发 并将类似的分析平台商业化。Advaita的现有产品目前由顶级负责人使用 调查人员、核心设施和世界各地的制药公司。

PROJECT SUMMARY Together with the ability to measure genome-wide expression of millions of individual cells, single-cell technologies have also brought the challenge of translating such data into a better understanding of the underlying biological phenomena. Existing computational methods and software for single-cell data analysis have critical limitations related to scalability, accuracy, usability, and interpretation capabilities. The main goal of this project is to pioneer a new platform for the analysis of single-cell data that is capable of: i) accurately identifying cell types and their composition in complex tissues, ii) inferring cell developmental stages and pseudo- time trajectories, and iii) identifying cell-type-specific pathways and putative mechanisms in a phenotype comparison. The proposed platform will also be able to deconvolve bulk expression data to identify the cell type composition of each bulk sample. The significance of the proposed work lies in its potential to provide new methodologies for single-cell data analysis that far exceed the performance of current state-of-the-art techniques. The accurate deconvolution will also allow researchers to extract more information from the vast repositories of existing bulk data, including GDC/TCGA, NCBI SRA, GEO, and ArrayExpress, which are currently containing data from bulk experiments that collectively cost over a billion dollars. The hypothesis driving this work is that single-cell data analysis and cellular deconvolution of bulk data can greatly benefit from: i) the systems-level knowledge that holds key characteristics for cellular developments, and ii) the valuable information available in validated cell types and reference single-cell datasets available in single-cell atlases. Indeed, our preliminary work shows that single-cell data analysis and cellular deconvolution can achieve an outstanding accuracy of approximately 90—100% if we properly utilize reference single-cell datasets and pathway knowledge. The proposed platform will be extensively validated by comparing its capabilities against the state-of-the-art software in both single-cell data analysis (cell type identification, developmental states and time-trajectory inference, systems-level analysis) and cellular deconvolution of bulk expression data. This will be done using both 663 datasets representing 279 cell types and 116 human organ parts (including bulk data, single-cell data, and matched cell flow cytometry). The pathway analysis and mechanisms inference capabilities will be further validated using real knock-out datasets (in which the true cause of the phenotype is known). The company, Advaita, has a strong IP portfolio, an experienced team, and a proven track record in this area, having developed and commercialized similar analysis platforms. Advaita's existing products are currently used by top principal investigators, core facilities, and pharmaceutical companies around the world.