The Cure: Making a Game of Gene Selection for Breast Cancer Survival Prediction

The Cure functions as a crowdsourcing platform designed to harness human intelligence for the identification of gene signatures critical to breast cancer survival prediction. By transforming complex genomic feature selection into an interactive gaming experience, the project aims to bridge the gap between massive biological datasets and the need for refined predictive models. Between September 2012 and September 2013, the initiative engaged over 1,000 participants who completed nearly 10,000 game sessions, utilizing 3,731 distinct genes derived from copy number variation and expression data.

The analytical framework relies on the WEKA software suite, which provides real-time feedback to players through C4.5 decision trees and 10-fold cross-validation. This methodology successfully captures expert-level biological knowledge, as evidenced by the significant enrichment of cancer-related diseases within the player-selected gene sets. These sets demonstrate predictive performance comparable to established commercial tests and existing literature-based predictors, confirming that aggregated human input can effectively navigate complex genomic feature spaces.

While the project validates the potential of games with a purpose to assist in biomedical research, the findings indicate that current iterations do not yet achieve statistically superior accuracy compared to traditional methods. The study concludes that while the platform successfully mobilizes a diverse and motivated population, future development must prioritize refined game design. Enhancing the user experience to reduce the learning curve and facilitating deeper exploration of the gene feature space remain essential steps for evolving these crowdsourcing tools into more robust instruments for clinical survival classification.