Multivariate Statistics and Machine Learning for Quality Control of Dried Ocimum Products

用于干罗勒产品质量控制的多元统计和机器学习

• 批准号: 10676412
• 负责人: Evelyn Abraham
• 金额: $ 4.01万
• 依托单位: PENNSYLVANIA STATE UNIVERSITY, THE
• 依托单位国家: 美国
• 财政年份: 2023
• 资助国家: 美国
• 起止时间: 2023-05-01 至 2025-04-30
• 项目状态: 未结题
• 来源: https://reporter.nih.gov/project-details/10676412
• 关键词: Anti-Bacterial Agents; Antioxidants; Attention; Bar Codes; Biological Assay; Biological Models; Botanical dietary supplements; Botanicals; Chemicals; Chemistry; Classification; Creativeness; DNA; Data; Data Collection; Data Set; Detection; Development; Dryness; Ensure; Flavonoids; Future; Genetic; Genetic Fingerprintings; Goals; Health; Health Benefit; Health Promotion; Herbal Medicine; Herbal supplement; Industry; Investigation; Least-Squares Analysis; Machine Learning; Marketing; Medicinal Plants; Medicine; Methods; Mission; Modeling; Modernization; Molecular Conformation; Multiomic Data; National Center for Complementary and Integrative Health; Natural Products; Ocimum; Ocimum basilicum; Output; Phytochemical; Plants; Property; Protocols documentation; Quality Control; Recommendation; Research Personnel; Resolution; Risk; Safety; Sampling; Scheme; Statistical Models; System; Testing; Therapeutic; Toxicology; Training; Validation; Variant; Visualization; Voting; antimicrobial; biomarker discovery; cell growth; cytotoxic; cytotoxicity; data modeling; drug development; genomic locus; improved; in vivo; machine learning algorithm; machine learning model; metabolomics; methicillin resistant Staphylococcus aureus; model building; model organism; molecular marker; neoplastic cell; novel; novel strategies; random forest; small molecule; statistics; tool; trustworthiness

PROJECT SUMMARY/ABSTRACT As the demand for medicinal plant products increases, so does the possibility of adulteration. Authentication of botanicals is complicated due to the immense quantity of molecular markers, including genetic loci and small molecules, within plant systems. This complexity also hinders identification of bioactive compounds responsible for the desired medicinal outputs. However, the improved accessibility of advanced statistical processing allows harnessing of these species-specific markers for sample identification and biomarker discovery. The overall hypothesis of this study is that multivariate and machine learning models will streamline multifaceted natural product investigations. Aim 1 applies multivariate statistics to genetic barcoding and high-resolution metabolomics data to develop authentication schemes, with Ocimum spp. (basil) as a model system. Random Forest and Partial Least Squares models are built using greenhouse grown, authenticated basil plants and used to predict the identity of consumer available products. Aim 2 uses the same statistical approaches to identify compounds responsible for both basil’s cytotoxic and antimicrobial properties. Developed models will also be used to predict dual-action bioactivity status of unknown samples. Models with the combined ability to identify bioactive compounds and samples will be recommended for future studies to improve compound discovery and classification of bioactive plants. The collection of data, development of statistical models, and professional development activities described herein will result in the development of a well-rounded, independent researcher.

项目总结/摘要 随着对药用植物产品需求的增加，掺假的可能性也在增加。认证 植物学是复杂的，由于大量的分子标记，包括遗传基因座和小 分子，在植物系统中。这种复杂性也阻碍了生物活性化合物的鉴定 用于所需的药物输出。然而，高级统计处理的可访问性得到了改善， 利用这些物种特异性标志物进行样品鉴定和生物标志物发现。整体 这项研究的假设是，多变量和机器学习模型将简化多方面的 天然产品调查。AIM 1将多元统计应用于遗传条形码和高分辨率 代谢组学数据开发认证方案，罗勒属。（罗勒）作为一个模型系统。随机 森林和偏最小二乘模型是使用温室种植的、经过认证的罗勒植物构建的，并使用 来预测消费者可用产品的身份。目标2使用相同的统计方法来识别 罗勒的细胞毒性和抗菌特性的化合物。开发的模型还将在 用于预测未知样品的双作用生物活性状态。模型具有识别 生物活性化合物和样品将被推荐用于未来的研究，以改善化合物的发现， 生物活性植物的分类。数据收集、统计模型开发和专业 本文所述的开发活动将导致开发一个全面的，独立的， 研究员