Write a comment

This publication aims to develop a platform of methods to identify small molecule ligand and RNA substrate interactions. Using a library of coumarin derivatives, they discovered a small molecule, C30, with high affinity binding to RNA single G bulges over other bulges (A, U, and C). The authors used Gaussian accelerated Molecular Dynamics (GaMD) simulations to study interactions, NMR to confirm the binding site location, and lasso regression to identify molecular descriptors for structure-activity relationships. The other narrative in this paper argues that this methodology could be used to develop better RNA-small molecule ligands for therapeutic purposes.

Major Revisions: 

1. To strengthen the novel aspects of the authors' methodology, it would be beneficial to discuss in a few sentences in the introduction why this method pathway is more advantageous than previous workflow paths. This addition would provide more clarity and depth to the methodological contribution. To drive this home in the conclusion, it would be beneficial to add some more sentences on the novel nature of this workflow, and how it was used to discover these new ligands. 

1. We suggest clarifying and justifying the type of RNA substrate used in each assay (ssRNA, dsRNA, etc.). For example, we suggest clarifying if the modelled version of “SL5RNA” used in the Fig 2 in vitro assay is the same as the additional FP simulations. For Fig. 3, in these FP simulations, please clarify if this “DNA version” of the RNA5 and RNA1 substrains is dsDNA or ssDNA. If it is helical dsDNA version of the substrate, a justification as to why this was used as a way to probe a minor groove binding mechanism for a seemingly bulged ssRNA substrate would be beneficial. 

1. We would suggest integrating some of the supplementary figures, like Supplementary Figure 6, into the main text. This could enhance the reader's understanding of the electrostatic interactions in RNA-ligand binding and the significance of Ring A's positive charge in identifying binding pockets.

1.  Additionally, we suggest exploring alternative feature selection methods such as SHAP or Markov trees. These could potentially capture more nuanced, nonlinear interactions that might be missed by a linear selection method like LASSO. 

Minor Revisions: 

1. In the introduction, we suggest that the authors clarify the importance of preventing deep hydrophobic binding pockets in a pharmaceutical context. We suggest a more streamlined discussion of coumarin derivatives and their therapeutic potential for SARS-CoV-2. 

2. We suggest the authors add “computationally” to “synthesized a collection…” for clarification in the text caption for Fig. 1A. We also suggest describing the principles and significance of using the fluorescence polarization assay to better justify the methodology. 

1. We suggest in Fig. 2D that the the graphs’ y-axes should be scaled the same and these curves should be labeled for “GA-rich” and “G” for clarity and not just the affinity values. 

1. For Fig. 3A, we suggest the legend be enhanced and add a scale bar for clarity. For 3D, the yellow, grey, and orange lines are not quite stark enough to differentiate. If there is significance to the groove being highlighted yellow, we also suggest listing this in the caption. 

1. If 4E is meant to show the same helical substrate as 3D, we suggest orienting them in the same direction and to better highlight differences. We suggest labeling “C12” for uniformity in 4E.

Competing interests

The authors declare that they have no competing interests.