查看文章

Single-cell technology offers unprecedented insight into the molecular landscape of individual cell and is transforming precision medicine. Key to the effective use of single-cell data for disease understanding is the analysis of such information through bioinformatics methods. In this thesis, we examine and address several challenges in single-cell bioinformatics methods for precision medicine. While most of current single-cell analytical tools employ statistical and machine learning methods, deep learning technology has gained tremendous success in computer science. Combined with ensemble learning, this further improve model performance. Through a review article (Cao et al., 2020), we share recent key developments in this area and their contribution to bioinformatics research. Bioinformatics tools often use simulation data to assess proposed methodologies, but evaluation of the quality of single-cell RNA-sequencing (scRNA-seq) data simulation tools is lacking. We develop a comprehensive framework, SimBench (Cao et al., 2021), that examines a range of aspects from data properties to the ability to maintain biological signals, scalability, and applicability. While individual patient understanding is the key to precision medicine, there is little consensus on the best ways to compress complex single-cell data into summary statistics that represent each individual. We present scFeatures (Cao et al., 2022b), an approach that creates interpretable molecular representations for individuals. Finally, in a case study using multiple COVID-19 scRNA-seq data, we utilise scFeatures to generate molecular characterisations of individuals and …