Shapify: Paths to SARS-CoV-2 frameshifting pseudoknot

Figures

• 

  Fig 1. SARS-CoV-2 viral genome. −1 PRF site marked by the red arrow. Normally, ribosome translates the complete ORF1a. Sometimes at the −1 PRF site, ribosome shifts from ORF1a to ORF1b, resulting in synthesis of fusion ORF1a/1b polypeptide.

• 

  Fig 2. SARS-CoV-2 −1 PRF pseudoknot native structure. Stem 1 in red, Stem 2 in black, and pseudoknotted Stem 3 in blue. Visualization generated using VARNA [11]. Mutations of interest shown with red arrows.

• 

  Fig 3. Pseudoknot diagram. A: Hairpin loop with stem base pairs shown in red. H-type pseudoknot pairing shown in blue between bases in the loop and complementary downstream bases. B: Representation of 3A in which arcs represent base pairs (arc diagram) for display of overlapping structures.

• 

  Fig 4. Initial stem 2 suboptimal structure. Predicted structures for the SARS-CoV-2 −1 PRF stimulating pseudoknot with initial Stem 2 (in red) as constraint. Blue arcs correspond with the minimum free energy structure, black arcs represent a suboptimal structure with only slightly higher free energy. Visualization generated using R-chie web server [71].

• 

  Fig 5. Cross validation results. For each possible combination of the intercept and slope, the value within the grid represents the geometric mean of the sensitivity and positive predictive value over 30 values of results from 29 RNA (each of the 30 RNA left out exactly once). Final result is the average of averages using a leave-one out scheme to avoid bias toward any one RNA. Optimal parameters identified by white box: intercept −0.5, slope 1.4. Color indicates performance relative to optimal with darker green as the best and red as the worst performance.

• 

  Fig 6. Shapify predicted native-adjacent structural paths. Presentation of structural paths from each initial stem leading to native or non-native (but native-adjacent) secondary structures (e.g., NN1 is the lowest free energy non-native structure; cf. Table O in S1 File). Initial stems labeled on the left (e.g., S1 for initial stem 1; cf. Table 4). If the structure predicted for a specific initial stem was the same for all four SHAPE datasets it is presented with a black colored path. In other cases, where the predicted structure was the same for three, two, or only one of the SHAPE datasets, the path is colored dark grey, grey, or light grey, respectively. Differences from the native structure are marked in bold, with parentheses/brackets representing changes in paired bases, and asterisks representing predicted unpaired bases that were paired in the native structure.

• 

  Fig 7. Shapify predicted native non-adjacent structural paths. Presentation of structural paths from the initial stems leading to native non-adjacent secondary structures. Initial stems are labeled on the left (cf. Table 4). If, for a given initial stem, the predicted structure is the same for all four SHAPE datasets, it is presented with a black colored path. In other cases, where the predicted structure is the same for three, two, or only one of the SHAPE datasets, the path is colored dark grey, grey, or light grey, respectively. Differences from native structure are marked in bold, with parentheses/brackets representing changes in paired bases, and asterisks representing predicted unpaired bases that are paired in the native structure.

• 

  Fig 8. Comparison of free energy of ShapeKnots and Shapify structural predictions. Each dot represents a unique secondary structure prediction for SARS-CoV-2 −1 PRF frameshifting pseudoknot. The y-axis represents free energy in kcal/mol with lower free energy corresponding with more energetically favourable structures. ShapeKnots (blue) failed to predict the most energetically favourable structures when given SHAPE data and initial pseudoknot free stems as constraint as compared to Shapify (orange). All structures are listed in Tables O-P in S1 File.

• 

  Fig 9. SHAPE dataset analysis. A: Comparison of −1 PRF sequence SHAPE reactivity dataset reported by Manfredonia et al. [13], Huston et al. [12], Yang et al. [46], and Zhang et al. [15]. Reactivity at or below 0.3 is considered to be low or non-reactive indicating the base is paired. B: ShapeKnots’ predictions using bootstrapped SHAPE values were obtained 10,000 times and averaged for each index. The mean and variance of the 10,000 predictions for each respective SHAPE dataset are shown here. The y-axis indicates the frequency each nucleotide is predicted as paired. Mean value of 1.0 at a specific position conveys that the nucleotide was predicted as paired for all bootstrapped values, and mean value of 0.0 indicates that the nucleotide was predicted as unpaired for all bootstrapped data. C: Bootstrap procedure was repeated with Shapify used for prediction. Red arrows mark differences from ShapeKnots predictions.