Generation of Riboswitch Design Statistics

  • 1
  • Idea
  • Updated 2 years ago
Hey guys,

I've been working on a high-throughput method of analyzing a riboswitch structure and sequence to generate a list of comparable statistics to determine if there are any global features that describe the behavior of a 'good' riboswitch, which could then be used as predictive statistics. Right now, I've managed to implement the ability to decompose a riboswitch into a series of functional domains, the interactions between domains, as well as general domain order for exclusion FMN switches.

Currently, I'm working to try and determine a general "distance" that individual bases move as the structures switch, and some other statistics I'm working on include:

-How many predicted base pairs break and form in each state, and at the smaller scale, for the individual helices. This includes information about the type and number of pairs, as well as helix length.

-Are there any correlations between base pair/helix free energies and the ability to switch?

-implementation of statistics for a 9-model system

So as a call out to the community, are there other interesting statistics that you can think of that may be generated from the predicted secondary structure or sequence? Also, if somebody knows how to mine the Eterna dataset for the predicted native structures for all parameters and states, that would be greatly appreciated.
Photo of Brourd

Brourd

  • 442 Posts
  • 79 Reply Likes

Posted 2 years ago

  • 1
Photo of Omei Turnbull

Omei Turnbull, Player Developer

  • 968 Posts
  • 304 Reply Likes
Couple of thoughts:

I think coaxial stacking of adjacent helices in multiloops plays a big role in our small-molecule switches, and is not adequately considered in the the energy models we have available.  So the existence of helices separated by 0 or 1 bases would probably be a feature worth considering.

On the topic of getting predicted foldings for all of our designs, LFP6 wrote a script for use by the new lab browser that loaded the flash game app and used its API to generate dot bracket notations for all states of all puzzles for a project.  I have yet to incorporate those results into the browser's database, but it certainly appeared to be generating what you want.   I've mis-placed the script ID at the moment; your best bet would be to contact him directly.
Photo of Brourd

Brourd

  • 442 Posts
  • 79 Reply Likes
Thanks for the heads up about generation of dot-bracket notation!

As for the snt separation, what kind of statistics are you looking for? I have some ideas, including:

-counting the number of helices that have these features

-a count of the number of helices that may switch from a structure with greater separation of helices to a smaller separation (this one may take a bit more time I would imagine)

-Some kind of helical length comparison for these features. My thought would be that if I can find a db with a collection of co-axial stacked helices, I can try to do a comparative search, or possibly create some kind of scoring system, which can then be compared against the predicted helices in our designs.

What other thoughts do you have?