A design principle for a single-stranded RNA genome that replicates with less double-strand formation

Usui, Kimihito; Ichihashi, Norikazu; Yomo, Tetsuya

The secondary structure predictions of RNAs shown in Figure 2 , and mdv-␣ mutants were prepared by Vienna RNA web services (http://rna.tbi.univie.ac.at/) (32,33) using the centroid structures. For viral genome analysis, we obtained all viral genome sequences registered in the NCBI database on 8 January 2015 (the list of accession numbers is provided in the Supplemental text), and the centroid secondary structures were predicted using the RNAfold algorithm in the 2.1.8 Vienna RNA package (32, 33) . We applied this analysis to sequences of all types of genomes (ssRNA, dsRNA, ssDNA, dsDNA and retrovirus genomes). Based on the secondary structural predictions, we counted the average loop size and average GC number in loops. For genomes larger than 4500 nt, we separated the sequences to a set of fragments less than 4500 nt, and after secondary structure prediction of the fragments, the GC numbers in loops of all of the separated fragments were averaged. Extremely large genomes (>40 000 nt) of some bacterial dsDNA viruses were removed before analysis to reduce the calculation load. The numbers of virus genomes in each category are shown in Supplementary Table S3 .