Multimodal AI for Multi-omics Data Integration

用于多组学数据集成的多模态 AI

• 批准号: 2870234
• 金额: --
• 依托单位: Queen Mary University of London
• 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-2870234
• 关键词: Multimodal; AI; Multi; omics; Data

Due to advances in genomics, transcriptomics, imaging and other modalities, health datasets are becoming increasingly rich, capturing detailed information about an individual's genotype and phenotype. Data from different modalities may provide complementary information, as well as exhibit correlated and/or causal relationships. But how to best combine multi-omics data for holistic prediction? This PhD will explore novel machine and deep learning solutions that integrate multi-omics data, with application to drug discovery science. Using neurodegenerative diseases like ALS, Alzheimer's and Parkinson's as exemplar domains, you will research AI data modelling and fusion methods to recognise patterns across multiple modalities, addressing challenges such as handling several omics modalities, data imbalances and missing modalities/features. Expected outcomes include new general purpose AI tools widely applicable to multi-omics data along with application of these tools in the discovery of new targets and biomarkers.The project will benefit from a number of emerging high-dimensional human datasets in neurodegenerative conditions that amalgamate transcriptomics, genetics, imaging and proteomic data and link them to disease phenotypes (e.g. AMP-AD, PPMI and Answer ALS). This presents an opportunity to explore the state-of-the-art multimodal integration approaches to define molecular underpinning on these conditions to define causal biology of the disease and ultimately leading to new ways of improving health in these debilitating conditions

由于基因组学、转录组学、成像和其他模式的进步，健康数据集变得越来越丰富，捕获有关个体基因型和表型的详细信息。来自不同模式的数据可以提供补充信息，并且表现出相关和/或因果关系。但如何最好地结合多组学数据进行整体预测？该博士将探索整合多组学数据的新型机器和深度学习解决方案，并将其应用于药物发现科学。使用 ALS、阿尔茨海默氏症和帕金森氏症等神经退行性疾病作为示例领域，您将研究 AI 数据建模和融合方法，以识别跨多种模式的模式，解决处理多种组学模式、数据不平衡和缺失模式/特征等挑战。预期成果包括广泛适用于多组学数据的新通用人工智能工具，以及这些工具在发现新靶标和生物标志物中的应用。该项目将受益于神经退行性疾病中的许多新兴高维人类数据集，这些数据集融合了转录组学、遗传学、成像和蛋白质组数据，并将它们与疾病表型联系起来（例如 AMP-AD、PPMI 和 Answer） 肌萎缩侧索硬化症）。这提供了一个探索最先进的多模式整合方法的机会，以定义这些条件下的分子基础，从而定义疾病的因果生物学，并最终导致在这些衰弱条件下改善健康的新方法