Scan statistics analysis for detection of introns in time-course tiling array data.

Area of Science:

• Genomics
• Bioinformatics
• Statistical Genetics

Background:

• Tiling arrays provide genome-wide abundance measurements using evenly spaced probes.
• Scanning window statistics are commonly used but can lead to biased interval predictions due to false discoveries and non-discoveries.

Purpose of the Study:

• To develop a novel statistical approach for analyzing tiling array data that minimizes false discoveries and improves detection power.
• To compare the performance of a proposed gene-wise search method against a traditional genome-wise scan.

Main Methods:

• A gene-wise search strategy was developed, considering at most one window statistic per defined genomic region (e.g., gene).
• This method was compared to a genome-wise scan that identifies clumps of adjacent probe discoveries.
• Simulations were conducted to evaluate false discovery rate (FDR) and statistical power.

Main Results:

• The gene-wise search method maintained the nominal false discovery rate (FDR).
• The genome-wise scan exceeded the nominal FDR for low interval effects and showed slightly lower power.
• In yeast, the gene-wise approach identified 71% of known introns and nine novel introns without false discoveries.

Conclusions:

• The gene-wise search strategy offers a more powerful and accurate method for analyzing tiling array data, particularly for sequence detection within defined genomic regions.
• This approach effectively reduces biases associated with traditional window-based methods and demonstrates practical utility in identifying genomic features like introns.