Novel Multi-Omics Integration and Machine Learning Tools to Understand Heterogeneous Systems of Ovarian Cancers

我的工作重点是将CT成像数据（放射组学）与代谢组学、转录组学和基因组学数据相结合。卵巢癌最大的挑战之一是化疗耐药性。我们的目标是通过各种组学透镜来研究卵巢癌，从而提高目前对卵巢癌的临床认识。具体来说，我们对多组学分析如何预测疾病结果感兴趣。在这个项目中，计划是用自上而下的表型到基因组学的方法来研究卵巢癌。我们收集了一个全新的、多地点、多组学的卵巢癌队列，促进了这一点。与此同时，我们正在开发各种预处理和机器学习工具，以提高放射组学、转录组学、基因组学和代谢组学结果的可重复性和有效性。一个大型的多组学卵巢癌队列结合一个强大的分析，反过来，将产生全面的深刻的结果；从临床医生到生物信息学家。在生物学水平上，我们将收集各种卵巢癌和非癌组织部位的组学数据。这包括卵巢癌组织样本和血浆样本的生化分析。除了这些人体样本，我们还获得了相应的时间序列增强对比CT扫描。我们试图了解组织分子病因学是如何反映在血浆组学档案，放射组学档案。这个新的多组学卵巢癌队列将采用传统的机器学习和深度学习技术进行分析。专门为这种类型的多组学队列设计的新型机器学习和深度学习方法也将得到开发。我们计划推广卵巢癌放射组学分析的“金标准”，以及多组学整合的新框架。总体目标是确定各种指示疾病恶性和化疗耐药性的制造者。最终，我们希望卵巢癌患者在正确的时间得到最好的治疗。

My work focuses on the integration of CT imaging data (radiomics) with metabonomics, trancriptomics and genomics data. One of the biggest challenges with ovarian cancer is chemotherapy resistance. We aim to enhance the current clinical understanding of ovarian cancer by investigating this disease through various omics lenses. Specifically, we are interested in how multi-omics profiles can predict disease outcome. In this project, the plan is to investigate ovarian cancer with a top down phenotype-to-genomics approach. This is facilitated by our collection of a completely novel a multi-locale, and multi-omics ovarian cancer cohort. In parallel with this lab component, we are developing a variety of pre-processing and machine learning tools to enhance the reproducibility and validity of radiomics, transcriptomics, genomics and metabonomics results. A large multi-omics ovarian cancer cohort combined with a robust analysis, in turn, will yield insightful results across the board; from clinic to bioinformatician.On a biological level, we will collect omics data from various ovarian cancerous and non-cancerous tissue sites. This includes biochemical analysis of ovarian cancer tissue samples and plasma samples. Along with these human samples, we also have obtained corresponding time-series enhanced contrast CT scans. We seek to understand how tissue molecular aetiology is reflected in plasma omics profiles, and radiomics profiles. This novel multi-omics ovarian cancer cohort will be analysed by employing traditional machine learning and deep learning techniques. Novel machine learning and deep learning approaches, that are specifically designed for this type of multi-omics cohort will also be developed. We plan to promote a 'gold standard' radiomics analysis of ovarian cancer, and a novel framework for multi-omics integration. The overarching aim is to identify a variety of makers that are indicative of disease malignancy and chemotherapy resistance. Ultimately, we want ovarian cancer patients to get the best available therapy at the correct time.