Yunyun Zhou

We focus on three primary research areas:

1. Machine Learning and Deep Learning methods for Driver Variant and Gene Discovery. We develop advanced deep learning methods, including Transformer models and Generative Adversarial Networks, to enhance the interpretation of genetic variants with unknown functions for cancer patients in clinical settings. This area also includes the creation of machine learning predictive models for cancer patient subtype stratification, survival outcome prediction, and drug treatment response prediction. In addition, we devise statistical feature selection methods and classification models for discovering and validating biomarkers in clinical cancer research. By linking the associations of genotypes and phenotypes, our goal is to better understand the genetic underpinnings of cancer and inform optimized treatment strategies.
2. Natural language processing for EHR Clinical Notes and Literature Mining. We utilize cutting-edge natural language processing (NLP) techniques, such as BERT and GPT, to identify psychiatric and behavioral phenotypes in clinical notes and literature. By leveraging real-world evidence, our goal is to improve early diagnosis and stratify neurodevelopmental disorder patients based on their distinct phenotypes. We have developed representation learning models for recognizing medical concepts and extracting phenotypes from temporal patterns in clinical notes. Furthermore, we have employed NLP knowledge graph techniques for knowledge discovery, connecting molecular variations with phenotypes and pathway alterations, which expedites knowledge acquisition across various domains, including COVID-19.
3. Integrative Network Approaches for Multi-omics and Clinical Outcomes Association Studies. We develop multi-layer network methods for predicting disease driver genes and functions by integrating mutations, genetic/epigenetic interactions, and functional pathways. This integrative approach allows us to gain insights into the complex interactions driving cancer development and progression. To support this research, we also create user-friendly cancer databases equipped with analytical tools that facilitate complex, dynamic, and high-dimensional cancer research. Our work enables both our team and fellow researchers to efficiently explore and analyze cancer-related data for enhanced understanding.