Computational biology and bioinformatics

Computational Biology and Bioinformatics is the application and development of computational methods to solve important problems in biology and biotechnology. At USJ, we develop novel machine learning models to learn the hidden patterns from biological data (DNA and proteins) and correlate them with the observed physiochemical properties and biological activities of these molecules. These models help us predict and design molecules with improved functions for drug development. In addition, our research involves modelling and simulating three-dimensional structures of proteins, ligands, membranes, and their assembly to gain insights into their interactions and functions. 

Computer-Aided Drug Discovery

Principal Investigator: Shirley W. I. Siu

Our research projects cover the early stages of drug discovery, from target identification to hit discovery and drug property evaluation. We have worked on both small molecules and short peptides as drug candidates, and have developed computational tools for novel drug discovery.

Antimicrobial Peptides

Antimicrobial peptides (AMPs) are a valuable source of antimicrobial agents and a potential solution to the problem of multidrug resistance. In particular, short-length AMPs have been shown to have enhanced antimicrobial activity, higher stability, and lower toxicity to human cells. To support the discovery and development of effective AMPs, we are developing machine learning and deep learning methods to recognize amino acid patterns of AMPs using publicly available data. Although relatively good classification can be achieved, it is still challenging to accurately predict the antibacterial assay value of peptides. The following figure summarizes the machine learning and deep learning approaches to predict AMPs.

For more information, see our review paper Recent Progress in the Discovery and Design of Antimicrobial Peptides using Traditional Machine Learning and Deep Learning, Antibiotics 2022.

Anticancer Peptides

Cancer is the major public health problem and the leading cause of death worldwide. A recent systematic analysis of cancer burden and trends in about 200 countries and territories found that the number of new cancer cases reached a new high of 23.6 million in 2019, and the number of cancer-related deaths was almost 10 million, equivalent to one in six deaths. Compared to the figures of 2010, this represented an increase of 26.3%  in new cases and 20.9% in deaths. Traditional chemotherapeutic agents kill cancer cells by directly interfering their DNA, RNA or protein synthesis, resulting in cell death. However, because of their non-specificity, they are cytotoxic to both normal and neoplastic cells, and these treatments often cause severe side effects or even lead to life-threatening conditions. Anticancer peptides (ACPs) are a potential alternative to current anticancer drugs. They are a subset of antimicrobial peptides (AMPs), which are naturally produced by the host’s innate immune system in response to invasion by pathogenic microorganisms. Our goal is to develop a computational workflow to accelerate ACP discovery by integrating efficient machine learning and deep learning models for classification, regression (“xDeep-AcPEP: Deep Learning Method for Anticancer Peptide Activity Prediction Based on Convolutional Neural Network and Multitask Learning”, see J Chem Inf 2021), and toxicity predictions.