Machine Learning Routines for Nanopore Sensing

纳米孔传感的机器学习例程

• 批准号: 2883765
• 金额: --
• 依托单位: University of Leeds
• 依托单位国家: 英国
• 项目类别: Studentship
• 财政年份: 2023
• 资助国家: 英国
• 起止时间: 2023 至 无数据
• 项目状态: 未结题
• 来源: https://gtr.ukri.org/projects?ref=studentship-2883765
• 关键词: Machine; Learning; Routines; Nanopore; Sensing

Nanopore technology is being successfully deployed for the sequencing of nucleic acids however it is still challenging to analyse heterogenous biomolecular samples. The analysis of datasets generated from nanopore sensors relies on a signal processing chain that employs advanced algorithms to classify translocation events to specific analytes passing through the nanopore. Often, the signatures of complex analyte mixtures are too convoluted to be analysed with high precision with current data analysis protocols. This project will develop of a fit-for-purpose data analytics approach to enable high-precision real-time analysis of highly convoluted nanopore datasets. While some features can readily be quantified using analytical tools (such as peak amplitude and dwell time), others (such as shape) are more challenging and will be better suited for analysis with machine learning approaches. The project will implement machine learning approaches for the clustering and classification of single molecule biosensing datasets generated using nanopores and functional DNA origami to support the development of the next generation of medical diagnostic devices.The project will also involve the development of multimodal characterization of catalytic nanoparticle systems where nanopore sensing will be complemented with electrochemical characterization at the single entity level. The project will implement sensor fusion algorithms to generate improved signal classification that will allow the development and characterization of new materials for a low-carbon future.

纳米孔技术正在成功地用于核酸测序，但分析异质生物分子样品仍然具有挑战性。从纳米孔传感器生成的数据集的分析依赖于信号处理链，该信号处理链采用先进的算法将易位事件分类为通过纳米孔的特定分析物。通常，复杂的分析物混合物的特征太复杂，无法用当前的数据分析方案进行高精度分析。该项目将开发一种适合目的的数据分析方法，以实现对高度复杂的纳米孔数据集的高精度实时分析。虽然一些特征可以使用分析工具（如峰值幅度和停留时间）轻松量化，但其他特征（如形状）更具挑战性，更适合使用机器学习方法进行分析。该项目将实施机器学习方法，对使用纳米孔和功能性DNA折纸生成的单分子生物传感数据集进行聚类和分类，以支持下一代医疗诊断设备的开发。该项目还将涉及催化纳米颗粒系统的多模态表征的开发，其中纳米孔传感将与单个实体水平的电化学表征相补充。该项目将实施传感器融合算法，以生成改进的信号分类，从而为低碳未来开发和表征新材料。