2020 Summer Scholars
At the MiCM, we’re embracing the “new normal” due to the COVID-19 pandemic by shifting in-person experiences online and working together to find creative solutions to new challenges.
We have kept our Summer Scholars Program and we are pleased to announce the selected 2020 Scholars. This summer, the 7 scholars will work under the supervision of an academic researcher on projects which span areas including chronic diseases, obesity, regenerative biomaterials, among others. In addition, the program manager, Dr. Amadou Barry, will be a resource person to assist scholars for any computational, statistical or other technical issues. We’re thrilled to have such talented scholars, and we thank you in advance for your hard work and creative thinking. You can find out more about our scholars as well as the projects they will be working on below.
See their progress over the course of the summer by watching their videos!
About the Summer Scholars Program
Launched in Summer 2019, the Summer Scholars program was developed to provide undergraduate and graduate students with the opportunity to learn about computational medicine through research projects. Scholars work with academic researchers, including the project PI and a senior data scientist, within the McGill community on a project for a period of 12 or 16 weeks in the summer months. More information on the program can be found here.
Dr. Amadou Barry
Summer Scholars Program Manager
- Heterogeneity of axon-myelin relation in gene
- Development of Agent-based simulation models for intervertebral disc biomaterials
- Data curation and exploring pre-treatment PET/CT scans using deep learning algorithm
- Profiling molecular steps in adipose tissue that lead to metabolic dysfunction in weight gain
- HLA Epitope Compatibility: A Strategy to Assess Risk and Personalize Transplant Care
- Sex and gender-related factors in chronic noncommunicable diseases
- Integrating machine learning applied to risk prediction of stress-related mental disorders
Myelin sheaths confer multiple properties on axons including vastly accelerated rates of action potential conduction with much reduced energy requirements. The profound functional significance of myelin is revealed by the severe neurological deficits that are consequent upon multiple inherited or acquired demyelinating conditions. While the dimensions of myelin sheaths relative to the variable calibers of the axons they ensheath approximate a constant ratio, recent observations demonstrate that this g-ratio can be modulated in response to adult experience with significant functional consequences. To further expose the basic rules underlying axon-myelin relationships, a computational approach will be applied to do so. That will involve the use of cable theory and the analysis of g-ratio and its effects on action potential propagation along axons. Heterogeneity of myelin sheath along axons and between axons will be considered.
Professor Anmar Khadra
Department of Physiology
anmar.khadra [at] mcgill.ca
Summer Scholar: Kusha Sareen
I’m an undergraduate student in Physiology and Mathematics going into my second year. I’m passionate about applying quantitative methods to biology and using computer science to solve physiological problems. In my free time, I like to read and play basketball.
In this project, Kusha will use a variation of the cable equation to study the propagation of action potentials (APs) along myelinated axons. He will study how heterogeneity in the thickness and length of internodes, as well as the radius of axonal fibers filter electrical signals of a given frequency and amplitude. The conduction velocity of APs will be also analyzed to quantify its dependence of the three previously listed anatomical parameters.
Kusha will participate regularly on weekly lab meetings where he is expected to present every 2-3 weeks. The supervisor will also meet him individually every week to discuss progress. Towards the end of the internship, Kusha will submit a report summarizing his finding with the expectation that the results produced will be eventually included in a more general manuscript focusing on myelin-sheath plasticity.
Intervertebral disc (IVD) disorder is the leading cause of lower back pain, affecting over 80% of the population and costing an estimated CAD $6-12 billion each year in Canada. Advanced IVD diseases often require surgical interventions, which have been associated with post-surgical complications and high long-term failure rates. Hence, there is high demand for new biomaterials and therapies that instruct cellular activities and healing processes. This necessitates the development of a computational platform that accelerates the invention and translation of novel bio-instructive biomaterials for the treatment of IVD disorders. Our goal is to integrate computer simulations and machine learning into the pipeline for tissue engineering research.
Professor Nicole Li-Jessen
School of Communication Sciences and Disorders
nicole.li [at] mcgill.ca
Voice and Upper Airway Research Lab
Summer Scholar: Grace Yu
I am a graduating student from McGill with an undergraduate degree in Physiology and Mathematics. Because I am interested in both biomedical and quantitative sciences, I am particularly passionate about research that connects the two fields.
The overarching purpose of the project is to use a combination of mechanistic modeling and machine learning techniques to predict the performance of IVD biomaterials, reducing the need for in vitro or in vivo experiments. Specifically, I will work on adapting Dr. Li-Jessen’s lab’s agent-based models of biomaterials (ABM-B) to simulate biomaterials for IVD repair. The expansion of the ABM-B will include the addition of cell types (IVD cells), material choices (chitosan), and tissue defects.
Because the biological reactions of IVD repair are highly complex and dynamic, I will then work with collaborators developing multi-task network-guided machine learning models to screen biomaterial design parameters, determining their importance to mechanical and biological properties of the biomaterials. With empirical experiments from our collaborators’ laboratories, targeted biomaterial designs will be evaluated, and the ABM-B will be iteratively optimized and verified.
The main goal is to develop the first generation of an agent-based model for IVD biomaterial design, which will be deposited on GitHub for storage and sharing with collaborators. Deliverables will also include weekly written reports of progress, and a final written report that is approximately 12 pages long (double-spaced with font size of 12), excluding the bibliography and appendix. Furthermore, a verbal presentation that is approximately 15 minutes long will be given at Dr. Li-Jessen’s team meeting.
Data curation and exploring pre-treatment PET/CT scans using deep learning algorithm and their outcome prediction modelling
Head and Neck (H&N) cancer is a group of neoplasms originating from the squamous cells that line the mucosal surfaces of the oral cavity, paranasal sinuses, pharynx or larynx. These cancer patients undergoing radiation therapy receive both anatomical CT and metabolic imaging (PET/CT). The prognosis of any individual patient is still often poorly determined. We recognize that from the tremendous amount of data generated by PET and CT imaging when a patient receives cancer care, much of which is currently not systematically used to its full potential. For our proposed project, we will further develop a novel deep learning framework that can take pre/during-treatment PET/CT images, H&N cancer patients’ clinical data and treatment planning radiation dose distributions, as inputs to augment the prediction of H&N cancer patient outcomes. We hypothesize that our architecture involving a novel training methodology where both PET and CT image branches are trained independently prior to being merged with a clinical data branch will obtain results superior than the current state-of-the-art models with only CT data incorporated as inputs. We will add to these images of radiotherapy dose distributions with the goal to augment prediction capability. Furthermore, we will also evaluate the impact missing data has on our model to determine its robustness. Therefore, we hope our model can accurately discern between high-risk patients and low-risk patients which will eventually lead to personalized treatments in the future. This project serves an incredible opportunity to combine my passion in machine learning and cancer research.
Professor Jan Seuntjens
Department of Medical Physics
jan.seuntjens [at] mcgill.ca
514-934-1934 X 44124
Summer Scholar: Yujing Zou
I recently finished U3 where I pursued a joint major in physiology and mathematics with a minor in physics. Starting Fall 2020, I will begin my master’s in Medical Physics at McGill. Throughout my undergraduate degree, I have been inspired and drawn to interdisciplinary research where mathematical modeling and computational tools are used to uncover non-linear dynamics problems in biology, as well as using machine learning and Radiomics to investigate various diagnostics imaging modalities in medical physics. For example, prior to joining medical physics research, I investigated a type of discrete mathematical model called Cellular Automata (CA) model and compared its predictability in real-time excitable media cardiac cell wave propagation data with a continuous model called FitzHugh-Nagumo (FHN) model. A Convolutional Neural Network (CNN) model integrated with the CA is currently being worked on to increase the model’s predictability at higher time steps. Furthermore, I have become interested in and dived into Radiomics, a field examining correlations between diagnostic image features and treatment outcomes in radiation oncology, since summer of 2018. There, I compared lymphadenopathy outcome prediction power between 2D (i.e. central tumor slice image) and 3D (i.e. whole tumor volume images) features from head and neck cancer dual-energy CTs (DECT) data at 21 energy levels from 87 patients’ DECT scans, and understanding how energy levels benefit different feature types.
Outside of academics, I am extremely passionate about writing, recording, and producing music and performing, as well as learning new programming tricks and languages.
- Under REB protocol, to collect clinical information data from computer servers for Head & Neck patients treated at the MUHC
- To curate the PET/CT imaging data and verify target and organ at risk contouring
- To extract radiomic features from the images, select the relevant features and analyze the correction with local control and distant metastases
- Analyzing these data using inhouse deep learning algorithms for testing the novel CNN model.
- We have a working deep learning model to make outcomes predictions for head and neck cancer patients. This model has been based on data from 4 hospitals in the Montreal area. We are now testing the model using data from a 5th hospital. A challenge has been to curate sufficient data for eligible patients that includes the necessary PET/CT information. The scholar will be key to help us extract and curate the data and work with a clinical medical physicist and a research assistant to build the DL model.
- A manuscript draft will be written once we have been able to confirm the model using the missing test set
- We will expand the model further by taking into account dose distribution information. This would the basis of a second paper.
- The scholar will make a presentation of her findings in MPU's Friday morning session.
Adipose tissue plays a central role in maintenance of healthy metabolism. In obesity, its capacity to expand via extracellular matrix remodeling, cell growth and lipid storage fails. This leads to ectopic lipid accumulation, whole-body low-grade inflammation, insulin resistance and ultimately to type 2 diabetes. This project examines the detailed profile of the extracellular matrix and inflammatory changes in adipose tissue that disrupt normal expansion during weight gain. The project uses whole adipose tissue transcriptome data set from a rare monozygotic, weight discordant twins. Focus will be on adipose tissue matrisome alterations and changes in inflammatory profile during adipose tissue expansion. Ultimate goal is to identify molecular targets that would allow maintenance of healthy metabolism in obesity.
Professor Mari Kaartinen
Faculty of Medicine (Experimental Medicine ) / Faculty of Dentistry
mari.kaartinen [at] mcgill.ca
(514) 398-7203 ext. 089668
Summer Scholar: Anny Hang
I am a student going into my last year of undergraduate as a Computer Science and Biology major. I am interested in the application of biotechnology in the medical field. In particular, I hope to later contribute in creating tools to facilitate the healthcare community.
The tasks involve making a program that organizes transcriptome data of the adipose tissue matrisome into heatmaps and find correlation between the change in composition of the extracellular matrix of adipose tissue, obesity and metabolic parameters associated with metabolic dysfunction.
A written report on the project (background, rationale, methods, results and discussion) and a presentation to the group.
Project description coming soon.
Professor Ruth Sapir-Pichhadze
Department of Medicine, Division of Experimental Medicine
ruth.sapir-pichhadze [at] mcgill.ca
(514) 934-1934 ext. 35403
Professor Yi Yang
Department of Mathematics and Statistics
yi.yang6 [at] mcgill.ca
(514) 398-4400 ext. 2793
Summer Scholar: Jacob Shkrob
My name is Jacob Shkrob. I’m an incoming junior at McGill University majoring in Honours Statistics and Probability with a minor in Computer Science. I’m also interested in genetics, biostatistics, and epidemiology. I’m from the south side of Chicago, IL, near the University of Chicago, and moved to Montreal to study at McGill. I’ve become particularly fond of mathematical analysis and algebra, and I’m keen on learning more about uses of Bayesian statistics and statistical inference in health care research. I’m also a musician and love to play classical music on the piano, as well as a poet and writer.
The project proposed for the scholar has 2 general parts:
The first part will involve working on the development and utility evaluation of machine learning methods for the analysis of cross-sectional and longitudinal health data. The proposed project for this part is UK Biobank. UK Biobank is a large prospective, population-based cohort study that was conducted via 22 assessment centers across England, Scotland, and Wales between March 2006 and December 2010 and recruited 502,624 participants aged 37 to 73 years. Over the study period, testing for COVID-19 in England was conducted and the data was provided by Public Health England (PHE) and linked to UK Biobank baseline data for further studies. The objective of this project is to establish risk factors of contracting COVID-19 infection in confirmed COVID-19 patients vs control using machine learning techniques.
The second part of the project will be integrating sex and gender dimensions in applied health research, to evaluate their impact on clinical cost-sensitive outcomes and patient reported outcomes related to quality of life in noncommunicable diseases including cardiovascular disease, metabolic disease, chronic kidney disease and neurological disease. We also aim to construct innovative ways to disseminate the application of gender measurement towards personalized approaches to chronic disease prevention, diagnosis and treatment in different countries. Some of these approaches are development and utility evaluation of synthetic data and exploration of federated analysis methods to enable multiple sites to perform a distributed analysis.
Professor Louise Pilote
Division of Clinical Epidemiology and General Internal Medicine
McGill University Health Centre Research Institute
louise.pilote [at] mcgill.ca
514 934-1934 ext. 44722
Professor Khaled El Emam
Faculty of Medicine, University of Ottawa
CHEO Research Institute
kelemam [at] ehealthinformation.ca
Summer Scholar: Yumika Shiba
I am an incoming junior doing Joint Honours Computer Science and Biology. I was born to biochemists, one of whom is also a molecular biologist, so I tried my best to avoid the “Bio-” path, which obviously failed. I worked with data from UK Biobank last summer, and I’m excited to work with it again, this time, examining it from a different perspective. I like reading and playing the ukulele for my free time.
- Establishment of risk factors of contracting COVID19 infection in confirmed COVID-19 patients vs control using machine learning techniques.
- Manipulation and Statistical Analysis of UK Biobank data.
- Development and utility evaluation of synthetic data.
- Exploration of federated analysis.
i. A report to Going FWD and MICM group
ii. PowerPoint presentation of study findings - presented to Going FWD meetings
iii. 1 manuscript
iv. Abstract and free communication
Project description coming soon.
Professor Xiangfei Meng
Department of Psychiatry
xiangfei.meng [at] mcgill.ca
514-761-6131 ext. 2352
Professor Yue Li
School of Computer Science
yueli [at] cs.mcgill.ca
+1-514-398-4400 ext. 2793
Summer Scholar: Michelle Wang