RiboSNitches — A Way to Explain Adverse Drug Reactions

Mukundh Murthy
8 min readNov 22, 2019

If you haven’t already read my article on noncoding genetic variants, look at that here. I give a basic introduction to the types of noncoding variants that can affect drug response, such as Expression Quantitative Trait Loci.

You probably remember the golden snitch from all of those Harry Potter Quidditch games. An impossibly fast golden sphere with wings that's hard for anyone to catch and determines who wins the game.

Well, riboSNitches might be the “golden snitches” of our genome, buzzing away from us as we try to understand the mechanisms behind drug response.

One nucleotide variant. That’s all it sometimes takes to change someone’s drug response and result in an adverse drug reaction.

Out of the 10,000 genetic risk variants identified in Genome-Wide association studies, most of the variants map outside of the protein-coding regions of the genome. These variants are called noncoding variants, and I’ve recently looked in riboSNitches, a class of single nucleotide polymorphisms that are characterized by changes in the three-dimensional structure of RNA.

Normally, we think of RNA being a single-stranded nucleotide sequence. However, these nucleotide bases on opposite ends of the strand can pair up, thereby causing the RNA to fold in complex ways.

tRNA normally folds up into an L-shaped 3d structure

The way in which an RNA is folded can greatly influence the ways in which it is regulated, processed, degraded.

A mutation in the nucleotide sequence can, therefore, change the 3D structure of the RNA (remember that A and T’s form two hydrogen bonds to each other and G and C form three hydrogen bonds to each other) and affect how it binds to certain essential RNA Binding Proteins (RBPs).

Here’s an example.

Ferritin Light Chain and Iron Responsive Element Binding Protein

Although we know that a riboSNitch will change the tertiary structure of the RNA, RNA with a given nucleotide sequence can actually adopt multiple conformations or shapes. Therefore, scientists use a complex technique called Principal Component Analysis (PCA) to map a multi-dimensional space onto a 2D coordinate plane.

The red clusters represented the normally folded FTL RNA. Mutations that occur in the purple portion act as riboSNitches that change the secondary structure of the RNA. Clusters are obtained through a process called Boltzmann Suboptimal Ensampling.

In the first image (a), you can see that there are three distinct clusters of structures for the ferritin light chain (FTL), represented by the colors red, blue, and green. The cluster of red structures represents the “correct” structure of the ferritin light chain, which properly binds the IREBP. However, when the 22nd Uracil nucleotide is mutated to a Guanine, or when the 56th Adenine nucleotide is mutated to a Uracil, the FTL RNA changes shape to adopt structures that are part of the blue and green clusters.

There are two main ways in which a riboSNitch can affect how an RNA binds to a protein. Imagine that you have a sequence of red, blue, green, and yellow lego bricks. Now imagine that a specific protein fits perfectly into a groove created by a consecutive sequence of red, blue, and green bricks, arranged consecutively next to each other. In this case, any part of your lego structure containing a consecutive sequence of red, blue, and green bricks is called a sequence motif. If you changed a blue brick in a sequence motif to a green one, then the protein wouldn’t be able to bind.

But… there’s another less intuitive way that you could prevent the protein from binding. You’d have to remember that the red-blue-green lego brick sequence motif isn’t isolated. It surrounded all around by other lego bricks, and its position is stabilized by the positions of the bricks surrounding it. What if you took a larger blue brick close to the sequence motif and replaced it with a smaller red brick?

In this case, although the sequence motif remains intact, the overall structure would shift. In the beginning, the sequence motif remained in a groove, but now it might be present on a bulge, and since your protein could only bind in a groove, it’s unlikely that it would strongly bind to a bulge.

In the same way, the IREBP does not bind strongly to structures in the blue and green suboptimal clusters, since the local structures of these clusters are different. For example, the green cluster has a larger circular sequence of RNA while the blue and red clusters have a longer hairpin loop.


A research technique called SHAPE (which stands for Selective 2′ -hydroxyl acylation analyzed by primer extension) can validate these changes in structure that we observed above in vitro. In order to understand how this technique works, we’ll have to look closely at the structure of an RNA nucleotide.

SHAPE targets the 2' OH (hydroxyl), group

SHAPE targets the 2' OH group of a sugar called ribose that is present in all RNA nucleotides. More importantly, it specifically aims to identify the residues that are flexible, and not base-paired to another nucleotide

Think about it this way: If each nucleotide was represented by a magnet, and two magnets attract each other to form pairs, SHAPE would be a huge magnet that attracts all the singular magnets that don’t have a partner. The more the number of singular magnets, the stronger the reactivity of SHAPE.

Therefore, as the number of flexible residues increases, SHAPE reactivity increases.

However, changes in structure are measured through changes in base-pairing probabilities, which are changes in the probability that one nucleotide will pair up with another nucleotide on the sequence. If a certain sequence undergoes a structural change where a greater proportion of nucleotides base pair, few nucleotides will be flexible, and SHAPE reactivity will go down.

SHAPE is an extremely useful technique since not all SNPs result in a change in structure. Therefore, SHAPE probing can be used to validate whether a certain nucleotide variant is acting as a riboSNitch or not.

Here’s an example of a graph showing how SHAPE analysis is done. (a) Comparing the SHAPE reactivity of the U22G mutation and normal FTL RNA shows that U22G is, in fact, a riboSNitch due to changes in SHAPE between nucleotides 30 and 60 (more reactivity = more flexible = riboSNitch). On the other hand, the Guanine to Adenine mutation in the fourth position is not an evident riboSNitch. You can see that the magenta and black line graphs almost overlap with each other exactly.

Why do we care about riboSNitches?

Here’s the central dogma of biology: DNA → RNA → Proteins

We’re normally taught in high school the central dogma of biology: DNA → RNA → proteins. Although this is true superficially, there are so many complexities that are overlooked when using this model.

For example, as we’ve already seen traditional “single-stranded” mRNAs and tRNAs can fold into complex shapes. Why, though??

Well, these RNAs regulate complex pathways that oversee the transcription and translation of certain genes. Therefore there 3D shapes can affect a multitude of processes such as…

  • Polyadenylation: A set of 300–350 adenine nucleotides are tagged onto the mRNA. These adenine nucleotides act kind of like a *** break at the end of a section or chapter in a book. It tells the reader (in this case, the RNA polymerase) to stop reading.
  • Splicing: The coding region of the genome can be broken up into protein-coding exons and interrupting introns. Splicing is a form of post-transcriptional regulation
  • miRNA targeting: If you had a piece of paper with text on it, but someone’s hand was covering it, you wouldn’t be able to read it right? In the same way, miRNAs are single-stranded pieces of RNA processed from mRNAs. They act as the “hand” that covers the mRNA by binding to it and tagging it for degradation.
All the steps that aren’t revealed by the central dogma

This is why investigating riboSNitches, and more generally, structural elements of the genome could help us formulate new nucleotide-based therapeutics as well as explain side effects in drugs that vary from person to person! If the shape of specific RNA sequences vary from person to person, thereby affecting how they bind RNA-binding proteins and regulate vital cellular processes, it’s important that we take account of them when we design new drugs!

What I’m Working On

Adverse Drug Reactions (ADRs) have doubled over the past decade and the number of cancer patients is predicted to exponentially rise up till 2050. How can we solve this problem? Analyzing noncoding regulatory elements of our genomes might hold the answer.

I’ve been looking through various research papers that harbor large amounts of data on riboSNitches and RNA structural variants. In addition, I’ve been looking to pharmacogenetic databases such as PharmGKB to find pharmacogenomic variants that are present in these riboSNitch databases.

RiboSNitches that simultaneously act as pharmacogenetic variants and expression quantitative trait loci (eQTLs) are likely the ones that we should target during drug development.

Here’s what I’m planning to do:

  • Continue browsing through PharmGKB, riboSNitch databases, and the human metabolome database to look for significant structural variants
  • Use programs such as RNAsnp to predict how a change in structure variant leads to a change in RNA structure. Use Boltzmann Suboptimal ensampling to generate Principal Component Analysis graphs and compare and contrast structures.
  • Measure binding affinity between different conformational states of RNA and other proteins involved in pharmacokinetic and pharmacodynamic pathways (pathways involved in drug metabolism and efficacy) through Autodock and SeeSAR software.

In the end, I plan to integrate all of these steps to create a more generalized workflow that allows one to

  • determine whether a mutation in an RNA can be classified as a riboSNitch
  • determine whether the riboSNitch causes adverse drug effects

As the price of human genome sequencing goes down, it's becoming easier for people to get their genome sequenced; however, this reduction in cost must be paired with a significant increase in our overall understanding of how certain nucleotides contribute to disease and drug response.

By understanding the disease and drug response mechanism, we will not only save billions of dollars in research and development but also prevent countless losses of loved ones due to a lack of personalized medicine.

Research Paper Used for the first part of the article (and images): https://www.ncbi.nlm.nih.gov/pubmed/26115028

Thanks for reading! Feel free to check out my other articles on Medium and connect with me on LinkedIn!

If you’d like to discuss any of the topics above, I’d love to get in touch with you! (Send me an email at mukundh.murthy@icloud.com or message me on LinkedIn)

Please sign up for my monthly newsletter here if you’re interested in following my progress :)



Mukundh Murthy

Innovator passionate about the intersection between structural biology, machine learning, and chemiinformatics. Currently @ 99andbeyond.