Computational tools for uncovering hidden patterns of DNA methylation

Innovative technologies now allow us to probe the epigenome in more dimensions and at higher resolution than ever before. However, meaningful biological insights are challenging to uncover in these high-dimensional settings where classical statistics fail, and relevant signals can be masked by technical noise and systematic biases. To unlock the full potential of genome-scale epigenomic data, our group develops novel statistical frameworks and tailored computational tools. In this talk I will highlight our recent and ongoing methodological innovations for single-cell, cell-free, and multi-omic DNA methylation analyses.

Probabilistic topic modelling to ascertain tumour microenvironments in large-scale transcriptomic data

Although cancer cells are marked by well-established somatic mutations, a wide spectrum of molecular heterogeneity across cancers, unknown gene regulatory mechanisms and the downstream effects of causal mutations still pose a great challenge in understanding disease mechanisms. A part of the challenges can be tackled by characterizing unknown tumour microenvironments that promote cancer cells' evolutionary advantages over surrounding cells. Here, we investigate how cancer cells evade immune surveillance and exploit immune systems by falsely propagating disguising messages to neighbouring immune cells using single-cell spatial transcriptomic data. We hope to statistically monitor TME changes by "eavesdropping" on cell-cell communications that happened between cancer and immune cells. We propose a topic model-based approach to ascertain millions of cell-cell interaction patterns manifested in gene expression vectors of ten thousand genes. We hypothesize that novel pathogenic gene programs can be distilled by examining genes that substantially co-expressed between cancer and immune cells.

Integrating Multi-Omics and Cross-Domain Machine Learning for Biomedical Insights

The rapid expansion of high-throughput biological data presents both opportunities and challenges for disease classification and biomarker discovery. Artificial Intelligence (AI) and machine learning (ML) provide powerful tools to integrate diverse data types, enhancing predictive accuracy. In this talk, I will discuss how multi-omics integration—combining different types of data, such as genomics, proteomics, and immunohistochemistry—can improve cancer classification. I will also explore the application of transfer learning in microbiome research, where models trained on one dataset or domain can be adapted to new but related problems. By leveraging data from multiple sources and domains, these approaches improve modeling and predictions.

The Battle Against Antimicrobial Resistance: From Academic Innovation to Real-World Impact

While antimicrobial resistance poses a growing global threat, large pharmaceutical companies have largely withdrawn from antibiotic development due to limited financial incentives. This critical gap must be filled by academic researchers and small/medium enterprises (SMEs) working in concert. Our journey in this space spans from AI-powered discovery at the Genome Sciences Centre to high-throughput validation at the UBC Centre for Disease Control, where we have developed novel antimicrobial peptides (AMPs) targeting World Health Organization’s list of priority pathogens.

Traditional drug development requires decades and billions of dollars to bring human therapeutics to market – resources beyond the reach of most SMEs. Recognizing this challenge, we founded Amphoraxe Life Sciences Inc. with a strategic focus on veterinary applications, where development timelines are shorter and regulatory requirements more streamlined, allowing us to make immediate impact. This choice addresses two pressing challenges in animal health: growing restrictions on antibiotic use in animal husbandry and increasing antimicrobial resistance. By leveraging our integrated AI platform and rapid validation system, we are developing novel AMP-based alternatives to traditional antibiotics, offering a path to both better animal health and reduced antimicrobial resistance pressure.

In this seminar, I will share my journey in launching a company to commercialize academic discoveries while maintaining active collaboration with academic labs. I will discuss how this academic-industry partnership model provides a template for addressing the antimicrobial resistance crisis, demonstrating how innovative technologies can be translated into practical solutions through strategic market positioning and focused development approaches.

Personalized genomic medicine approaches to expand treatment options for cancer patients

British Columbia's Personalized OncoGenomics (POG) project aims to rationally identify cryptic cancer cell vulnerabilities and thereby expand treatment options for poor prognosis cancers. Ongoing since 2012, POG has relied heavily on comprehensive whole genome and transcriptome sequencing analysis and in-house data interpretation and reporting software to enrol more than 2,000 BC patients onto study and return results to oncologists and, through them, to patients. This work has been foundational to the Terry Fox Research Institute's Marathon of Hope Cancer Centres Network, which now aims to expand precision cancer genomic medicine to jurisdictions across Canada. With the well-established analytic backbone POG has built, it is now possible to begin evaluating other technologies and assess the extent to which these might further expand on our ability to inform clinical treatment planning and bring benefit to cancer patients and their families.

Abstract TBA!