Clustering with exclusion zones: genomic applications

Area of Science:

• Spatial and temporal statistics
• Genomic data analysis
• Statistical methodology

Background:

• Scan statistics are widely used for event clustering in space or time.
• These methods assume no "holes" (exclusion zones) in the data domain.
• Unknown exclusion zones complicate accurate statistical inference.

Purpose of the Study:

• To develop a method for formally evaluating event clustering in the presence of unknown exclusion zones.
• To create an algorithm for estimating the total extent of unknown exclusion zones.
• To enable bias correction for scan statistic-based inference in complex domains.

Main Methods:

• Developed an algorithm to estimate exclusion zone extent using "spacings" distributional properties.
• Implemented bias correction techniques for the exclusion zone estimator.
• Assessed algorithm performance through simulation studies.

Main Results:

• The proposed algorithm effectively estimates total exclusion zone extent.
• Bias correction is demonstrated to improve the accuracy of the estimator.
• Simulations confirm the algorithm's robust performance.
• Applications in genomics reveal significant changes in inference when accounting for exclusions.

Conclusions:

• The developed algorithm provides a robust solution for analyzing event clustering with unknown exclusion zones.
• Accurate estimation and correction for exclusion zones are critical for reliable statistical inference, particularly in genomic studies.
• This work advances the application of scan statistics in complex spatial and temporal domains.