This project seeks to advance personalized medicine through two critical research objectives, each focusing on early diagnosis using data-driven approaches.
The first objective is to identify the causal links between genetic variations (SNPs) and anomalies associated with brain function disorders such as Alzheimer’s disease, Parkinson’s disease, and autism. By integrating genotype data with imaging features, we aim to uncover imaging biomarkers that mediate the relationship between genotypes and phenotypes. Large-scale convolutional neural network (CNN) models, such as ResNet and VGG, will be used for analyzing imaging data, while traditional machine learning methods like logistic regression and support vector machines (SVM) will be employed to analyze SNP data. These analyses will leverage the UKBB dataset, focusing on association analysis between genotypes, imaging features, and clinical phenotypes of brain function disorders. This approach avoids the use of large language models or generative methods, adhering strictly to interpretable, hypothesis-driven methodologies. The ultimate goal is to develop a framework capable of identifying critical imaging biomarkers to assist in the diagnosis and understanding of brain function disorders, improving precision in early-stage diagnosis and treatment.
The second objective focuses on the early detection of pancreatic ductal adenocarcinoma (PDAC), a highly lethal cancer. By analyzing blood metabolite profiles, we aim to identify specific markers capable of distinguishing high-risk individuals (e.g., those with a family history of PDAC) from the general population before the onset of cancer. These metabolite markers will also have prognostic value for diagnosed PDAC patients. A diagnostic model will be developed to stratify high-risk groups and detect PDAC in its early stages, potentially enhancing survival rates.
