The Seeker is looking for a comprehensive approach – perhaps using some data mining tools involving classification and regression trees, cluster analysis, etc. - for each of two related problems. Both problems involve analyzing a wide range of covariates that may help (a) identify sub-groups of patients who have a differential response to treatment, and (b) predict responders and non-responders in clinical trials.
Since these approaches can be exploratory in nature, the Seeker is also interested in methodologies for validating predictive models derived from the data. An actual data set relevant to (b), which is the integration of 5 clinical studies, is provided (though anonymized and disguised for confidentiality reasons) for Solvers to test/demonstrate proposed methodologies.
The submissions to this challenge should include:
- A brief paper (5-10 pages) explaining the pros and cons of different predictive approaches and how they can be applied to clinical trials, along with a recommendation for a general methodology/approach that could be applied routinely in clinical trials. Consideration for both (a) and (b) as defined above is desirable.
- A report on the analysis of the integrated clinical trial data set that is provided for item (b), including descriptions of statistical tools and code (S+, R, SAS, etc.) used in the analysis. The report will also describe the interpretation of the analysis, conclusions and recommendations on how reliable the results are based on building and validating the predictive model for responders and non-responders.
The challenge award is contingent upon evaluation of the proposed methodology by the Seeker. The Seeker is not looking for a literature review. The proposed methodology should provide the Seeker with "freedom to practice," i.e. there should be no patents or patent applications which would prevent the use of the proposed solution in commercial applications.
Up to 4 awards ranging from $1,000 to $5,000 will be issued for the most useful solutions. Also, the Solvers offering the most interesting solutions may be offered by the Seeker continued collaboration with the intent to publish research papers in statistical and medical journals.