Hi Eterna,
I finally got around to writing my new RNA document and it is turning out to be a book… Im dumping everything I have learned and figured out over 10 years? of research on the partition function and the ensemble, as well as how RNA is best stabilized and whatnot. Im done with chapters 1 and 2 which teach the basics of the ensemble kinda like a ensemble 101 best practices and what is actually happening inside the ensemble to make the ensemble. This is so that I can next explain why I believe that RNA is more rules by mechanical rules than energy rules and such and that mechanics of materials is essential for proper RNA design… that will come in the following chapters…
Here is a link and a PDF… I would love comments… It is a draft and not spell checked very well if at all yet… I suck at that but you all know that…lol.
https://docs.google.com/document/d/1klD7haeAc4UYyKJQem2JzguMGYIrZHy9zc4Yw4SJVh0/edit?usp=sharing
A mechanical overview of RNA in an energy focused world - Draft rev1 (2).pdf (71.0 KB)
2 Likes
I like this so far! Here are my first impression comments:
- I thought you did a good job conveying that it’s important to consider the ensemble and not just the MFE
- I found the visual descriptors about nucleotides tugging on each other useful for picturing what’s happening in my mind
- In later drafts, having pictures/examples could be useful for conveying some of the ideas - whether those are things like cartoons or schematics of what the RNA structure is doing, or things like concrete plots/graphs/analyses from past labs, or something like an example of the subopt function in NUPACK
- Related to “To start with, I truly believe that RNA is a mechanical system that lives in an energy world, and thus us bound by the rules of mechanics of materials more than the effects of the energies in the loops.” - I’m really curious to hear any comments/analyses/hypotheses/etc. regarding the interplay between mechanics and the different folding engines in future chapters
I’m excited to read more whenever it’s ready!
Hi Everyone,
Thank you @Anamfija for your feedback. I incorporated it a bit into the next chapters. Here is a new draft in which I have connected teh RNA energy world and nucleotide bonds to the first 3 laws of physics as part of a connection to statics and then mechanics of materials. It also explains the ensemble and how the pairing probabilites form it, or at least it is the start of that. its at 19 pages now…
https://docs.google.com/document/d/1JLQkozfaPkzQaMH6t0PbTDCvU05bVOEvnqugAuFui8w/edit?usp=sharing
A mechanical overview of RNA in an energy focused world - Draft rev3-shared.pdf (2.0 MB)
1 Like
It would be helpful to visually represent the relationship among pairing probabilities, the ensemble, and the partition function. Maybe use a 20nt RNA as an example.
Are your pairing probabilities based on experimental data or strictly an algorithm? My understanding is that the pairing probabilities provided for Vienna2 and NuPACK are based on experimental data. Is that correct?
I like that you are writing a document for Eterna players to explain the concepts and pull together your observations, and the process might lead to a specific hypothesis or a specific topic to develop for a journal paper.
the pairing probabilities for the hairpin example are from in silico calculations on for a design I have not synthesized in vivo. When I talk about switchs and go further I will be using designs that have experimental data behind them. The purpose of this first part was to show what the calulations are predicting in the background and then to apply in chapter 5 and beyond with real world examples
Hi Jennifer, I added a few suggestions in your doc. Hopefully they help a bit. My first thoughts are that it’s doing a good job with the ensemble/MFE. The pictures were good, I hope more can be added later on to further explain what you are conveying. Just a few grammar mistakes, but those can be fixed later on easily. The big comment from me is probably the nature of how you are giving the information, particularly the POV and the formalness. I think once you decide fully want you want this to be (document, book, etc.), it can be changed to accommodate that nature. Overall, it’s very informative and I can’t wait for more to be added.
so i had mentioned that I had a theory that there is some quantum effects in the RNA… well here is article in the BBC talking about how DNA has been found to have quantum tunneling between the helix and that causes mutations of DNA… If it happens to DNA I doubt it does not happen to RNA as it is the protons that are responsible for the bonding (I think they mean hydrogen bonds) that have ben found to be tunneling between the entire energy barrier at both sides of the helix. I think this sounds like every binding site has quantum tunneling to every other binding site. I think that might go a long way to explaining the source of the pairing probability mechanical connection, and why RNA does RNA stuff… humm deep thoughts
https://www.sciencefocus.com/news/quantum-weirdness-could-be-the-driving-force-behind-dna-mutations/
I think that shape data from earlier labs might actually be perfrect for this. I need to investigate this further… If only I had this understanding when we were doing SHAPE labs… I think I finally understand them (better? and who really does) now that I have this info in my head…
I should clarify that when I asked if pairing probabilities are based on experimental data, I meant data external from Eterna. (Das Lab has not generated any ensemble data that I’m aware of.) I read today that pairing probabilities are based on the lowest MFE and the frequency with which they appear in an ensemble. The structures with the lowest MFEs usually appear the most frequently, but not always. My original assumption was that the pairing probabilities come from the research team at Vienna. Or if we are using NuPACK, those folks have put together the pairing probability data used by NuPACK.
Then I see both Vienna and NuPACK use McCaskill’s partition function algorithm for computing base pair probabilities. Are the base pairing probabilities within different RNA structures so regular (repeated) that an algorithm can capture (reliably reflect) all that data?
in relation to McCaskils algorith (m-fold) it is so far teh most efficent and accurate found for getting the QUICK partition function and thus MFE and ensemble. McCaskill is not more accurate than others and in fact he is less accurate at times but always really fast and very accurate for what you need. In fact the website for M fold actually says to be careful and not trust the MFE but to poke around and verify at different energy levels aka subopt which I pushed to implement. You can use a different algorithm that is similar to -d2 in Vienna and I always use that in Vienna adn would think that we should use that here as m fold verses -d2 is m fold is fast and not as accurate at times and -d2 is slow but very accurate and concise.
@DigitalEmbrace Would it be possible to do some experiments with the other 2 settings for the partition function? Currently we are using M-fold 2.? which allows varying temps entered. M-fold3.? is available but will only do 37C and there is the -d2 setting i mentioned above. I think that we just went with the default and there may be a better option as you alluded to in your messages I feel.
The fact that there are options for different temperatures does suggest that the partition function is based on experimental results. As for which setting to use, that would be something for Rhiju to decide.
@DigitalEmbrace
Here are a couple good links
https://www.researchgate.net/publication/258565726_MfoldC_RNA_modeling_program
this one actually has a list of advantages adn disadvantages The disadvantages speak to teh fact that we really should be using subopt more if we are using mfold folding as it is not “right all the time”. There are new models. We could probably run a fold through UnaFold or something that replaced m-fold and then run that fold through nupack for pknot. I think that would be doable for at least single state to get more accurate predictions with modern models.
http://rothlab.ucdavis.edu/genhelp/mfold.html
http://www.unafold.org/