Deep Learning Methods to Integrate Biological Information for Analysis of Single-cell RNAseq Data

项目摘要 该项目广泛的长期目标涉及新型机器的开发 由重要因素驱动的基因组数据建模的学习方法和计算工具 生物学问题和实验。单细胞RNAseq(ScRNAseq)数据分析 带来了巨大的计算和生物信息学挑战。该计划的具体目标是 项目是利用先验生物信息开发新的基于模型的深度学习方法 考虑对scRNAseq数据进行建模。这些问题都是由PI的收盘引发的 与生物医学研究人员的合作。建议的方法旨在 整合生物信息以提高分析性能和生物学性能 可解释性。这些方法依赖于生物学洞察力和深度学习的新整合 分析噪声、稀疏和过度分散的scRNAseq数据的方法，包括零- 膨胀的负二项模型、自动编码器、深度嵌入、双曲线嵌入以及 逆图嵌入。新的方法可以应用于两个重要的生物学 使用scRNAseq技术的问题：通过集群识别和发现细胞类型 通过轨迹推断进行分析和细胞发育。它们将有助于有效地分析 日益重要的scRNAseq数据集，并对正在进行的重要研究作出贡献 PI目前正在合作研究人类头发的潘氏细胞调节和再生 毛囊。该项目将开发实用可行的计算机程序，以便 实施所提出的方法，并通过以下方式评估这些方法的性能 真正的应用程序。本文提出的工作将有助于为建模提供深度学习方法 ScRNAseq数据，研究复杂的表型和生物系统，并提供见解 进入由各种数据集表示的每个生物区域。已开发的所有计划 根据这项赠款，感兴趣的人将免费获得详细的文件。 研究人员。来自不同背景的本科生研究人员将作为一个 项目中不可或缺的一部分，用于实施拟议目标的最关键部分。这个 研究项目将激发学生的兴趣，使他们能够考虑在 生物医学科学。

Project Summary The broad long-term objective of the project concerns the development of novel machine learning methods and computational tools for modeling genomic data motivated by important biological questions and experiments. The analysis of single-cell RNAseq (scRNAseq) data presents substantial computational and bioinformatics challenges. The specific aim of the project is to develop novel model-based deep learning methods with prior biological information considered for modelling scRNAseq data. These problems are all motivated by the PI’s close collaborations with biomedical investigators. The proposed approaches are designed to integrate biological information for improving both analytical performance and biological interpretability. The methods hinge on novel integration of biological insights and deep learning methods for analysis of the noisy, sparse, and over-dispersed scRNAseq data, including zero- inflated negative binominal model, autoencoder, deep embedding, hyperbolic embedding, and reversed graph embedding. The new methods can be applied to two important biological problems using the scRNAseq technologies: cell type identification and discovery via clustering analysis and cell developments via trajectory inference. They will facilitate effective analyses of the increasingly important scRNAseq data sets and contribute to the important on-going studies that the PI is currently collaborating on, Paneth cell regulation and regeneration of human hair follicles. The project will develop practical and feasible computer programs in order to implement the proposed methods, and to evaluate the performance of these methods through real applications. The work proposed here will contribute deep learning methods to modeling scRNAseq data and to studying complex phenotypes and biological systems and offer insights into each of the biological areas represented by the various data sets. All programs developed under this grant and detailed documentation will be made available free-of-charge to interested researchers. Undergraduates researchers from diverse backgrounds will be recruited as an integral part in the project for implementing most critical parts of the proposed aims. The research project will stimulate the interests of students so that they can consider a career in the biomedical sciences.