Albany 2015:Book of Abstracts

Albany 2015
Conversation 19
June 9-13 2015
©Adenine Press (2012)

Improving nucleic acid design using biased sequence initialization

Nucleic acid sequences can be designed to self-assemble to a desired conformation. They can be used as pharmaceuticals, to detect molecules or as nanomachines. Designing nucleic acids can be thought of as inverse of secondary structure prediction. One approach to design algorithms consists of three main steps: 1) sequence initialization, 2) structure decomposition and 3) refinement. Prior work in this field initialized the sequences using equal composition of nucleotides. We hypothesized that by using a biased sequence initialization instead of the default random approach, the time efficiency and accuracy of the design would be improved. We tested this hypothesis using ensemble defect optimization as pioneered by NUPACK (Zadeh et al. 2011).

First, we searched a grid over possible combinations of single and base pair compositions for sequence initialization, and we found a set of optimum parameters that had the best speed and accuracy for design. Second, we mined a database of known RNA secondary structures and used their average nucleotide content as a method for the initialization, inspired by evolution. Third, we designed sequences to a set of target structures, and mined the designed sequences to use their average nucleotide content for initialization. In all cases, the biased sequence initialization leads to better and faster designs. In the case of the grid search method, an average of 15 times speed up is found. Further improvements in speed are possible by implementing a GPU version of the partition function calculations of base pairing probability (Stern and Mathews 2013) in the refinement step.

    Stern, H. A. and D. H. Mathews (2013). Accelerating calculations of RNA secondary structure partition functions using GPUs. Algorithms Mol Biol 8(1): 29.

    Zadeh, J. N., B. R. Wolfe and N. A. Pierce (2011). Nucleic acid sequence design via efficient ensemble defect optimization. J Comput Chem 32(3): 439-452.

Mohammad Kayedkhordeh
Stanislav Bellaousov
David H. Mathews*

Department of Biochemistry and Biophysics
University of Rochester
601 Elmwood Avenue, Box 712
Rochester, NY 14642, USA

Phone: (585) 275-1734
Fax: (585)-275-6007