Most cancers lack effective markers to detect disease early, to identify high risk patients (prognostic signatures), and to predict optimal treatment strategy (predictive signatures). The main challenge is learning from and using the enormous amounts of complex, molecular data from tumor samples. Many methods have been introduced to generate prognostic and predictive signatures from these data, but we lack systematic approaches to find all useful signatures, and to select the best one.
Counter intuitively, good signatures comprise not only the most
differential markers. Identifying these less differential markers - "the best supporting actors" - calls for novel, integrative approaches as current statistical and machine learning methods fail in this task.
The Mapping Cancer Markers (MCM; http://www.cs.utoronto.ca/~juris/MCM.htm
project aims to comprehensively and systematically discover clinically
useful markers to help:
1. Detect cancer early,
2. Identify high-risk patients, and
3. Predict treatment response.
To power this research, we rely on World Community Grid volunteers who
donate their computers' spare capacity to carry out this extensive
analysis. Finding all clinically useful markers would require processing thousands of patient samples and testing an astronomical number of marker combinations, which is not feasible even on World Community Grid.
We developed approaches called heuristics to reduce the search space,
enabling us to tackle this challenge. However, the success of MCM project is only enabled by World Community Grid with the computing resources donated by volunteers around the world.
Considering the diversity of analyses and data, we will have large
variation in runtimes and work unit duration. We will keep you updated on the progress and discoveries, and would be happy to answer any questions you may have about the project.
We hope you will find our project interesting, and support its goals. With your participation, we will form the largest "cancer research team" in the world.