Machine learning of time-series single-cell drug screening to elucidate HIV latency control mechanisms

项目总结 基因表达动力学为潜在的结构-功能关系提供了丰富的见解 基因电路和疾病的蓝图。对于病毒来说尤其如此，这些病毒的决策 高度依赖于它们的基因表达动态。时间序列基因表达扰动数据 阐明生物因果关系，对于确定生物机制、进行化学 生物分析，以及新药的发现。然而，一条全面和 有效分析这类时间序列扰动的数据并不存在。在这项工作中，我们建议应用 机器学习揭开人类免疫缺陷病毒(HIV)时间的隐藏(潜在)结构-- 当受到化学干扰时，一系列基因的表达。这种高度跨学科的努力 包括与NIH任务相关的科学领域，如生物学、临床、物理、 化学、计算、工程和数学科学。建议的研究领域 将机器学习、病毒学、系统生物学、化学科学、单细胞生物物理学和 药学。该研究将培训和支持两名教职员工和两名研究生 为期两年的研究助理。

PROJECT SUMMARY Gene expression dynamics yield a wealth of insight into the underlying structure-function relationships of gene circuitry and the blueprints of disease. This is especially true for viruses whose decision-making is highly dependent on their gene expression dynamics. Time-series gene expression perturbation data elucidates biological causality and is essential to identify biological mechanisms, perform chemical biology analyses, and for the discovery of novel drugs. However, a pipeline to comprehensively and effectively analyze such time-series perturbation data does not exist. In this work, we propose to apply machine learning to unravel the hidden (latent) structure of human immunodeficiency virus (HIV) time- series gene expression when affected by chemical perturbations. This highly interdisciplinary effort includes scientific areas relevant to the mission of the NIH such as biological, clinical, physical, chemical, computational, engineering, and mathematical sciences. The proposed areas of research combine machine learning, virology, systems biology, chemical sciences, single-cell biophysics, and pharmaceutical sciences. The research will train and support two faculty members and two graduate research assistants for the two-year term.