Figure 1.

Procedure of encoding the RNA sequences and the amino acid sequences into feature vectors. a) Procedure of encoding the RNA sequences into feature vectors. For the secondary structure, RNAsubopt was used to obtain the top six possible secondary structures with the lowest free energy. The dots and brackets were then replaced by 0 s and 1 s, respectively. The six vectors were added, and the secondary structure feature vector was obtained. For Van der Waal’s interaction and hydrogen bonding, each base was replaced by numbers representing the propensities. Finally, all three feature vectors were transformed by the Fourier series, and the first 10 terms of Fourier series were used as the new feature vector. b) Procedure of encoding the amino acid sequences into feature vectors. For the feature vector of the secondary structure, the corresponding Chou-Fasman propensities were used to encode each amino acid according to the secondary structure predicted by Predator. For the feature vectors of hydrogen bonding, each amino acid was replaced by Grantham’s and Zimmerman’s scores, respectively. Kyte-Doolittle and Bull-Breese scores were used for Van der Waals’ interaction, respectively. For all five feature vectors, the first 10 terms of the Fourier series were used as new feature vectors.

Lu et al. BMC Genomics 2013 14:651   doi:10.1186/1471-2164-14-651
Download authors' original image