Project News

1. The systematic review of AI technique applications in genomic analysis, and 
2. The potential applications of generalized out-of-distribution (OOD), an emerging machine learning research concept, in the biomedical domain. 
    

Artificial Intelligence Applications in Biomedical Genomics

Authors: Kairui Guo, Mengjia Wu, Zelia Soo, Yue Yang, Jie Lu (UTS), Hua Lin, Mark Grosser (23 Stands).

Abstract: Genetic and genomic research has evolved to become a critical clinical tool for newborn screening, diagnostics of inherited diseases, predictive testing for complex diseases, and precision medicine based on pharmacogenomic findings. Machine learning has proven invaluable for annotating a broad range of genomic sequence elements and investigating underlying gene expression mechanisms. In recent years, artificial intelligence (AI) techniques leveraging genomic data have been developed and implemented in various clinical settings, exerting a direct influence on the work of clinicians.

Here, we present a systematic review of AI-assisted genomic applications in biomedicine, demonstrating how various AI algorithms are contributing to disease prevention, prediction, diagnosis, prognosis, and treatment. Additionally, several unresolved challenges, ranging from gene-specific model development to governance and regulatory management, are identified here. We aim to provide a timely review to computer scientists, biomedical researchers, and clinicians to familiarize themselves with the current trends in AI technologies and genomics, thereby equipping them with the knowledge necessary for future advancements in this exciting field.

Generalized Out-of-distribution Detection: Theory and Application

Author: Zhen Fang (UTS).

Abstract: Generalized out-of-distribution (OOD) detection is vital for ensuring the safety and reliability of artificial intelligence systems. It represents a novel and trending area in machine learning and artificial intelligence. The concept of generalized OOD detection was first proposed in 2017 and has since shown significant potential in enabling the reliable deployment of machine learning models in real-world applications, including medical safety and autonomous driving. Over the past few years, a rich line of algorithms has been developed to address the generalized OOD detection problem empirically. In this talk, we will present the latest advancements in theoretical understanding and the medical applications of generalized OOD detection.