Project 1 illustration

Deep learning models for understanding the non-coding genome

We are developing deep learning models to predict the impact of mutations in the DNA on cellular function across diverse cell types. We are also building interpretable models for understanding regulatory motifs that are predictive of cell-type specific activity of non-coding sequences.
People: Chendi Wang, Sasha Maslova
Alumni: Mark Ma
Project 1 illustration

Modeling the impact of genetic variation on multiple types of cellular traits

We are developing integrative computational approaches for predicting the impact of genetic variation on multiple types of cellular traits, including DNA methylation, histone modification,and gene expression.
People: Bernard Ng, Farnush Kiadeh
Alumni: Rosemary McCloskey, James Topham
Relevant publications: here and here
Project 2 illustration

Deconvolution of hidden and known confounding factors

We are developing machine learning methods for modeling the effect of hidden and "partially known" confounding factors in gene expression and methylation data. In this direction, we are currently developing methods for estimating cell type proportions in "genomics" samples from varied tissues including brain and cord blood.
Peaple: Louie Dinh
Relevant publications: here
Project 3 illustration

Multi omics analysis of complex disease

Working with clinical colleagues at UBC and elsewhere, we are developing association analysis approaches using multiple types of omics data. Specifically, we are currently focusing on integrative analysis of genotype, gene expression, DNA methylation, and histone modification data in order to identify genes and networks that are associated with neurodegenerative diseases (including AD & PD).
People: Bernard Ng, Halldor Thorhallsson
Project 4 illustration

Constructing tissue- and cell-type specific regulatory networks

We have developed machine learning approaches for constructing gene functional association and gene regulatory networks in a context- and tissue-specific manner. Expanding in this direction, we are interested in developing network construction algorithms that combine varied types of genomics data (in particular data from DNA accessibility assays such as ATAC-Seq, gene expression data, and protein expression data) to model gene regulatory networks in a cell type specific manner.
Alumni: Emma Pierson
Relevant publications: here, here, and here