What the lab is like

We maintain both wet (benchwork 🥽 🥼 🧫) and dry (computational 💻) labs, providing a very inter-disciplinary environment for our group members. Everyone is encouraged to do (at least a little bit of) both, and we are all committed to helping new members with this cross-training through both formal and ad hoc events.

We are committed 💯 to a diverse, equitable and inclusive environment where everyone can function at their absolute best 🌈 🏳️‍⚧️ . Please see our lab ethos for a more detailed description of what this means for us.

Much of our lab work is based on modern “-omics” 🧬 🧬 technologies. Currently, we develop single cell profiling technologies and lentiviral-based strategies for high-throughput CRISPR/Cas9-mediated perturbation delivery. So people with backgrounds in the life sciences and who enjoy lab work have opportunities with us.

The lab has significant expertise in computational biology and data science. We are particularly interested in the concept of expression signatures and expression networks (you can learn a bit about them below). Ultimately we are interested in applying these techniques to the data from our perturbation studies to build computational models that help us reason about cellular causal and correlative relationships. So people with quantitative backgrounds (computer science, mathematics, statistics) may enjoy developing probabilistic models and deep learning toolkits with us to tackle these problems.

The lab also co-leads efforts to profile large breast cancer cohorts. In particular, we are interested in early “pre-invasive” forms of breast cancer including ductal carcinoma in situ. The lab has a long-standing interest in breast cancer informatics where we apply our novel assays and models to better understand the disease at the molecular level, and where we build expression signatures that may have clinical utility in the treatment of breast disease. So people who are interested in clinical applications and translational research have opportunities with us.

We are a highly collaborative lab, participating in projects with different groups and institutions. Dr. Vanessa Dumeaux is one such long-term collaborator (she’s also my wife), and our labs share space and equipment; we often have joint lab meetings. The lab also organizes a yearly meeting in systems biology, giving us an excellent opportunity to interact with some of the best researchers in the world. We really love Do-It-Yourself (DIY) science, and are committed to Open Access Science.

You can learn more about our group, how it works and its dynamics here in this presentation. Afterwards, we provide slightly more technical sketches of the major directions in the lab. We note however that we are always open to good ideas from new members!

Development of new deep learning techniques in the life sciences

From a methods perspective, our primary interest is in the development of data science and machine learning techniques for the analysis of molecular data. We have a number of on-going computational projects in the lab, many of which are rooted in fundamental data science methodology (statistics, visualization, hypothesis testing) or deep learning. Most of these projects are motivated by, and support, the applications described below. Mike was a professor in the School for Computer Science at McGill for ~20 years and has experience supervising students from the quantitative sciences.

There are MSc and PhD positions available for individuals with strong backgrounds in the quantitative sciences to work on methods building related to the projects described below.

Systematically perturbing biological systems

Our lab has several projects related to the inference of biological networks. We have developed approaches that perturb specific sets of genes or transcripts using genome editing (eg CRISPR) in a high-throughput, multivariate pseudo-random manner. The edits can be knockouts, knock-downs or over-expression of the targets, or a combination thereof. Single cell high-throughput profiling of transcripts, select proteins, or chromosomal conformations are used to deduce underlying biological networks - the regulatory and correlative relationships between genes and gene products.

These projects bring together all the components of our lab:

• Single cell multi-modal -omic profiling;
• Genome editing;
• Next generation sequencing;
• Data science and analysis of high-throughput data; and
• Deep learning for inference of biological networks.

We apply this to different biological systems including in the context of breast cancer with Sylvie Mader’s group at the IRIC/UdeM in the past.

There are two graduate level positions (MSc, PhD) open. There are opportunities for people with strong mathematics, computational and statistical background. The work is primarily based on probabilistic modelling and deep learning focused on inferring biological networks from the data genearted by our assays. There are also opportunities for people with strong bench skills that would like to work with lentivirus-delivered CRISPR-mediated molecular perturbations. Please reach out if you are interested.

Identifying effective treatments for women with early breast disease

This is a CIHR funded project that is a collaboration with Dr. Eileen Rakovitch at the Sunnybrook Hospital in Toronto, Canada.

[Ductal in situ carcinoma (DCIS)(https://www.mayoclinic.org/diseases-conditions/dcis/symptoms-causes/syc-20371889) is a very common form of breast lesion. DCIS is a non-obligate precursor to invasive “life-threatening” breast cancer invasive ductal carcinoma (IDC). Specifically, not all DCIS will become invasive, if left untreated (at least not within the natural lifetime of an individual). In constrast to indolent DCIS, these lesions will progress, escaping from the mammary duct, into the surrounding breast tissue. Eventually, if left untreated, they metastases to other organs and tissues. A woman with an indolent DCIS might decide in consultation with her health care practioners tol follow a milder form of therapy involving only breast conserving surgery (BCS) However, if there is evidence she has a more agressive DCIS, then treatment might include BCS with additional radiotherapy. The fundamental problem is that we have no way currently to decide whether a woman with DCIS benefits from the additional radiotherapy: we do not have biomarkers to differentiate those DCIS likely to remain indolent from those that will progress to an invasive state. Therefor,e current clinical practice is cautious and leans towards the inclusion of radiotherapy.

Dr. Rakovitch is a radiotherapist who has had great foresight and energy to build a cohort tumor banking of DCIS cases in Ontario from 1993 until present. This cohort is very well annotated with patient, clinical and pathological information about the lesion and patient. We are interested in identifying a molecular signature that can predict whether a woman with DCIS would benefit from a regime with only BCS versus a harsher regime of BCS + raditiotherapy. To do this, we are profiling Dr. Rakovitch’s Ontario cohorts using next generation sequencing and analysis methods from computational biology. This data and analyses allow us to glimpse how each such lesion evolved, the collection of somatic DNA events in each tumor and the changes in the regulation and expression of specific pathways. From this data, we are developing molecular signatures that have the ability to predict at time of diagnosis the benefit from the inclusion of radiotherapy to assist clinical decision making.

We have positions open at all levels from undergraduate projects, graduate students, postdocs and research associates. Please reach out if you are interested.

Phenotypic heterogeneity in the fungus Candida albicans

We recently modified the DROP-seq protocol (a single cell sequencing method) from Macosko et. al. to function with the fungus C. albicans. The main differences between fungi and mammalian cells, where DROP-seq is well establsihed, are that the former have a cell wall that needs to be removed before lysis, and the latter have considerably more mRNA. Our system was used to profile C. albicans grown in the presence of different antifungal drugs. We observed heterogeneity in their response, with evidence of distinct subpopulations of survivors with differential survival responses bioRxiv preprint. There is many opportunties for the development of single-cell informatic techniques optimized for the fungal setting for students with quantitative backgrounds. In addition to the single cell profiling techniques, we have also been building deep learning tools for fungal microscopy Candescence. This is ongoing work.

This is joint work with PhD candidate Samira Massahi, Vanessa Dumeaux at Western, Judy Berman, and Malcolm Whiteway at Concordia.

Collaborative efforts related to breast cancer

We have many more projects related to breast cancer that are highly inter-disciplinary in nature. Our collaborators span across Canada from Dalhousie to UBC, to the United States including MD Anderson, NYU and Alabama, and in Europe with researchers in Norway, Sweden, Holland and Italy. We are always looking for individuals of all backgrounds that are intrested in cancer biology and would like to contribute to the data science underlying these studies.