Detecting genomic clustering of risk variants from sequence data: cases versus controls

Area of Science:

• Genetics
• Statistical Genetics
• Bioinformatics

Background:

• Dense genetic marker measurement is approaching DNA sequence limits.
• Clustering of genetic variants, especially rare ones in functional regions, can offer biological insights.
• Existing statistical tests for genetic variant clustering include spatial scan statistics.

Purpose of the Study:

• To extend the Tango's spatial clustering statistic for genomic sequence data.
• To compare the performance of the extended Tango's statistic (Kernel Distance) with other clustering methods, including the Kulldorff scan statistic and the sequence kernel association test.
• To evaluate Type-I error rates and statistical power across various clustering scenarios.

Main Methods:

• Extension of Tango's spatial clustering statistic to genomic sequence data, termed the "Kernel Distance" statistic.
• Comparison of the Kernel Distance statistic with the Kulldorff scan statistic and the sequence kernel association test.
• Evaluation of Type-I error rates and power under different genetic variant clustering scenarios.

Main Results:

• The Kernel Distance statistic, an extension of Tango's method, was computationally faster than the Kulldorff scan statistic.
• The Kernel Distance statistic exhibited slightly lower statistical power compared to the Kulldorff scan statistic.
• The Ionita-Laza version of the Kulldorff scan statistic demonstrated the highest power across the evaluated clustering scenarios.

Conclusions:

• The Kulldorff scan statistic is a powerful tool for identifying clustered genetic variants in association studies.
• Extending spatial clustering methods like Tango's can provide rapid computational alternatives, though power may be slightly reduced.
• These methods are crucial for leveraging dense genetic marker data and uncovering biological insights from variant clustering.