Grant: DRA : Canada Graduate Research Scholarship - Doctoral Program (CGRS D)
McGill University
$120,000.00
Incorporation of causal principles for credible predictions in precision oncology.
EF 2025-26
Résumé
Precision oncology is the idea that cancer treatment should be tailored to each person's unique cancer, instead of giving the same treatment to everyone. This approach has already led to successes in drug treatments that target specific cancer types. However, radiation therapy is still mostly given using a "one-size-fits-all" approach. Research in precision oncology is moving quickly, often using artificial intelligence and machine learning (ML) to find patterns in genetic, imaging, and clinical data. These methods could make cancer care more personalized, but progress has been slow. Many studies use small datasets, do not handle missing information properly, and rely on "black box" ML models that are hard for doctors to understand or trust. We believe these problems limit the use of new tools in real clinics. Our idea is that by applying causal principles (methods that study cause-and-effect relationships) we can make predictions that are more reliable, easier to understand, and more useful for patients.
This research has three main goals. First, we will find better predictors of treatment response, looking for biological and imaging features that can explain how tumours respond to radiotherapy. We will choose these predictors based on cause-and-effect reasoning, making sure they are clear and useful across different cancer types. Second, we will detect and reduce bias, building tools to measure and correct common problems in prediction models, such as bias from missing data or poor testing methods. These tools will be shared through free, open-source software. Third, we will build clearer ML models, including biological knowledge in their structure. This combines the power of AI with the clarity doctors need to trust the results. Overall, this project aims to create prediction tools that are trustworthy, understandable, and ready to guide future cancer treatment.
Programme
DRA : Canada Graduate Research Scholarship - Doctoral Program (CGRS D)
Bénéficiaire
Emplacement
Quebec, Canada
Type de bénéficiaire
Academia
Type
Grant
Source
CIHR Funding Decisions Database
Tel que publié
Id
560251
Name
Glickman Harry W
Region
Québec
Country
Canada
Orgtype
Research Institution; Institution Paid
Instname2
Cancer Research
Theme2
Biomedical
Prcname2
DT1 (Biomedical and/or Clinical Research) CGRS Doctoral
Projecttitle
Incorporation of causal principles for credible predictions in precision oncology.
Programtype2
Studentship Programs
Orgnameinp2
McGill University
Cihramount2
$120,000
Cihrequipment2
$0
Programname2
DRA : Canada Graduate Research Scholarship - Doctoral Program (CGRS D)
Keyworddelim
Bias; Biomarkers; Cancer; Deep Learning; Genetic Markers; Genetics; Machine Learning; Precision Oncology; Prediction Algorithms; Prognosis
Abstract
Precision oncology is the idea that cancer treatment should be tailored to each person's unique cancer, instead of giving the same treatment to everyone. This approach has already led to successes in drug treatments that target specific cancer types. However, radiation therapy is still mostly given using a "one-size-fits-all" approach. Research in precision oncology is moving quickly, often using artificial intelligence and machine learning (ML) to find patterns in genetic, imaging, and clinical data. These methods could make cancer care more personalized, but progress has been slow. Many studies use small datasets, do not handle missing information properly, and rely on "black box" ML models that are hard for doctors to understand or trust. We believe these problems limit the use of new tools in real clinics. Our idea is that by applying causal principles (methods that study cause-and-effect relationships) we can make predictions that are more reliable, easier to understand, and more useful for patients.
This research has three main goals. First, we will find better predictors of treatment response, looking for biological and imaging features that can explain how tumours respond to radiotherapy. We will choose these predictors based on cause-and-effect reasoning, making sure they are clear and useful across different cancer types. Second, we will detect and reduce bias, building tools to measure and correct common problems in prediction models, such as bias from missing data or poor testing methods. These tools will be shared through free, open-source software. Third, we will build clearer ML models, including biological knowledge in their structure. This combines the power of AI with the clarity doctors need to trust the results. Overall, this project aims to create prediction tools that are trustworthy, understandable, and ready to guide future cancer treatment.
Competitiondate
202511
Approvedterm2
3 yrs 0 mth
Pinamesdelim
Glickman, Harry W
Approvedterm
3 ans 0 mois
Cihrcontribution2
~ $120,000~ $0~