Theme 1: Deep omics!
Deep learning has revolutionized research in image processing and speech recognition and will soon transform research in molecular biomedicine. Deep learning models can capture multiple levels of representation directly from raw data without the need to carefully engineer features based on fine-tuned algorithmic approaches or domain expertise. Omics data is one of the most prominent examples of feature‐rich and high‐dimensional heterogeneous data and thus multi-omics data analysis and integration have increasingly become a deep learning harvesting field in computational biology. We are developing deep learning models to leverage large omics data for finding hidden structures within them, for integrating heterogeneous data and for making accurate predictions in different biomedical applications ranging from single-cell omics analysis and multi-omics biomarker discovery to human functional genomics and drug discovery.
Theme 2: Network biology and systems-based biomarker discovery
Recent advances in high-throughput technologies have provided a wealth of genomics, transcriptomics, and proteomics data to decipher disease mechanisms in a holistic and integrative manner. Such a plethora of -omics data has opened new avenues for translational medical research and has particularly facilitated the discovery of novel biomarkers for complex multi-factorial diseases (e.g., cancers, diabetes, neurodegenerative diseases). We have a number of collaborative projects on integrating multiple data sources, network and temporal information using advanced computational approaches to better understand the molecular complexity underpinning pathogenesis and to identify novel and precise biomarkers for disease early-detection, diagnosis, prognosis and drug responses paving the way for personalised medicine.
Theme 3: Computational Drug Repositioning
De novo drug discovery is an expensive and time-consuming process. During the past years, there has been a surge of interest in drug repositioning to find new uses for existing drugs. Repositioning is economically attractive when compared with the cost of de novo drug development; it can reduce the traditional timeline of 10-17 years and make drugs available for use in 3-12 years. The number of repositioning success stories is rapidly increasing, and more companies are scanning the existing pharmacopoeia for repositioning candidates.
Computational repositioning is an emerging multidisciplinary field to develop automated workflows that can generate hypotheses for new indications of a drug candidate using multitude of high dimensional molecular data. This project is aimed to use transcriptomics, drug-target interactions, and/or genome-wide association studies (GWAS) to systematically generate repurposing hypotheses for candidate drug molecules.