Statistical Analysis of RNA Backbone
Abstract
RNA backbone conformation analysis has been demonstrated to be particularly difficult due to the large number of torsion angles per residue and the large variability of the raw data. Due in part to the importance of local structures in the understanding of RNA catalysis and binding functions, studies in this area have recently received increased attention. In this work, the authors use classical tools from statistics and signal processing to search for clusters in the RNA backbone torsion angles. Results are reported both for scalar studies, where each torsion angle is separately studied, and for vectorial studies, where several angles are simultaneously clustered. Using techniques from optimal quantization, they automatically find the torsion angle clusters. With these clustering techniques, they find RNA backbone motifs, both at the single residue level (phosphate-to-phosphate) and at the suites level (base-to-base) parsing. These two parsing techniques also are compared using mutual information measurements. The authors conclude the work with statistical analyses of some of these motifs, and optimal fitting of torsion angle distributions in the most significant clusters. The whole process is fully automatic and based on well-defined optimality criteria.
Document Details
- Document Type
- Technical Report
- Publication Date
- Feb 01, 2004
- Accession Number
- ADA437808
Entities
People
- Allen Tannenbaum
- Eli Hershkovitz
- Guillermo Sapiro
- Loren D. Williams
Organizations
- University of Minnesota