Exploratory Data Mining for Subgroup Cohort Discoveries and Prioritization.

Area of Science:

• Computational Biology
• Biomedical Informatics
• Data Mining

Background:

• Identifying homogeneous subgroups within large, heterogeneous populations is crucial for biomedical research and hypothesis generation.
• Existing computational methods lack robust strategies for identifying novel and clinically relevant subgroups with well-designed studies.

Purpose of the Study:

• To develop a novel subgroup discovery method for identifying potential cohorts with explainable contrast patterns.
• To provide a data-driven framework for tailoring interventions in precision health.

Main Methods:

• Employed a deep exploratory mining process to analyze thousands of potential subpopulations.
• Prioritized cohorts based on explainable contrast patterns and interventionable insights.
• Conducted computational experiments on synthesized and clinical autism data, including scaling analysis.

Main Results:

• Demonstrated quantitative performance in covering pre-defined cohorts using synthesized data.
• Showcased qualitative novel knowledge discovery from a clinical autism dataset.
• Provided insights into computational resource needs for large-scale subpopulation analysis.

Conclusions:

• The developed method offers a robust framework for automated subgroup discovery.
• This approach facilitates the identification of novel hypotheses and clinically relevant cohorts.
• The findings support the advancement of precision health through data-driven intervention tailoring.