A novel scan statistics approach for clustering identification and comparison in binary genomic data.

Area of Science:

• Genomics
• Biostatistics
• Computational Biology

Background:

• Identifying specific genomic regions with high event rates is crucial in biomedical research.
• Traditional methods often require pre-defined cluster parameters, introducing arbitrariness.
• Spatial scan statistics offer a non-parametric alternative for detecting event clusters.

Purpose of the Study:

• To adapt the Bernoulli-model scan statistic for genomic applications.
• To develop a multivariate extension, Relative Scan Statistics, for comparing two Bernoulli random variable series.
• To identify unshared event rate variations and clusters within genomic data.

Main Methods:

• Adaptation of the Bernoulli-model scan statistic to the genomic field.
• Development of a multivariate extension (Relative Scan Statistics) for comparing two Bernoulli random variable series.
• Utilizing a probabilistic, likelihood-based hypothesis testing procedure for cluster identification.

Main Results:

• The proposed Relative Scan Statistics effectively identify genomic clusters and relative clusters.
• Both univariate and multivariate extensions confirm previously published findings.
• The method highlights unshared event rate variations between two series of Bernoulli random variables.

Conclusions:

• The scan statistics framework is effectively applied to genomic data analysis.
• This method enhances understanding of viral vector integration processes.
• Researchers can focus on specific genomic regions for targeted investigation.