Fig 1.
Schematic illustration of a potential well with a minimum in the known aptamer sequence.
It is possible to escape the minimum by performing nucleotide mutations to the initial sequence until there are no BLAST hits when inserting the mutated sequence as input. If the mutations disrupted the known aptamer structure, the use of RNAfbinv will restore the known structure while generating designed sequences as output. Subsequently, designed sequences in the borderline of the potential well that do show BLAST hits should then be carefully examined in their hits. It is expected that most of the hits observed will be from known bacteria but a few unknown bacteria and exceptional eukaryotic organisms will also show up.
Fig 2.
Starting from a known aptamer sequence, performing random nucleotide mutations to escape from the similarity of the known aptamer sequence, some of the mutations may disrupt the aptamer structure (See Fig 1), followed by directed mutations that restore its structure using RNAfbinv [70] and offer newly designed sequences for inspection. During these phases we generate multiple designed sequences. The designed sequences are BLASTed against large DNA databases such as NCBI's Nucleotide colletion nr/nt. Similar BLAST results are folded using the Vienna RNA package [57] and compared visually to the known aptamer secondary structure. Similar sequences are marked as potential aptamers. The special case of covariant mutations explores sequences that fold to the same structure, thereby not allowing for the flexibility of the fragment-based design.
Table 1.
Results obtained with the purine riboswitch: new bacterial aptamer domains detected by our method.
Fig 3.
Outputs of the search flow shown in Fig 2.
Selected outputs taken from the search, resulting in Fig 4. (a) Output screen for the multiple random mutation phase. On top, the initial known aptamer sequence–the xpt guanine-binding riboswitch aptamer, below are some of the generated sequences. (b) Output of multiple RNAfbinv runs. On top, the input including the starting sequence (marked in the frame above), below are designed sequences from multiple runs. (c) Output of nucleotide BLAST for input sequence of Run 7 (marked in the above frame), below are the BLAST matches including the three marked results that appear in Fig 4.
Fig 4.
Results using the purine riboswitch for method validation.
The predicted secondary structure drawings of three new bacterial aptamer domains detected by our method, in comparison to the known xpt guanine-binding aptamer domain. Shown in red are preserved nucleotides.
Fig 5.
A potential prokaryotic new finding.
The predicted secondary structure drawing of an additional new bacterial aptamer domain detected by our method, in comparison to the known xpt guanine-binding aptamer domain. The aptamer sequence does not align via BLAST with the xpt riboswitch. Shown in red are preserved nucleotides.
Fig 6.
A potential eukaryotic new finding.
A potential match found in Aspergillus oryzae, in comparison to the Mesoplasma florum riboswitch [75]. Essential nucleotides for a potential ligand binding site are marked in both secondary structures. Structure similarity between the two compared structures is relatively high: Base pair distance = 38, Shapiro distance = 4. Sequence similarity is relatively low but the essential nucleotides make it an attractive riboswitch candidate. Similar to [75], a 3D view of the ligand-binding pocket bound to its ligand is depicted on the right-hand side. This predicted purine riboswitch candidate found in fungi is not detected by Infernal.