Testing the pseudoknot: A tale of two methods

jandersonlee has written a new version of the mutation/submission booster that generates a unique barcode for our OpenKnot round 1 to facilitate mass submission of sequence variation. Players can add the booster to their in-puzzle tool list by going to Mutation booster with random barcodes and clicking the green Favorite button.

One way to use this new tool is to take a known pseudoknot and mutate a few bases to test if the pseudoknot still forms.

Begin by selecting a known pseudoknot from the UTEP database and pasting the sequence in the puzzle using the Sequence Stamper on the booster list (lightning bolt icon). I’m going to paste the tobacco virus pseudoknot at position 40.

Then paste the dot-bracket at the same position. (Right-click or control-click to bring up the menu.) The UTEP database uses semicolons instead of periods so you’ll need to change those to dots. (An easier way is to open one of the PKB designs mjt and I have already submitted with the custom target structure.)

Paste Secondary Structure

Now I can see the known structure of the pseudoknot. Notice how I have padded the surrounding sequence with AAAAAGAAAA to reduce the numbers of A in a row and to reduce the risk of the tobacco virus sequence misfolding with another portion of the puzzle. The tobacco virus sequence happens to have 7A in a row; I will leave those in place since that is the original sequence.

I want to test whether the pseudoknot still forms when a closing pair changes or no longer pairs. I use the Mark Bases tool to select the two bases.

Then open the Mutation Booster and click Mutate.

After the script finishes generating the sequences, click Submit. Enter a title for your design. The PKB ID is a helpful title so that we can track which pseudoknot is being mutated. Then enter a description such as “Testing closing pair”.

Lastly, sit back and watch the robot do its work.

1 Like

The other method that I have been using is to load a known pseudoknot with its known secondary structure and make specific changes to the sequence to test if the pseudoknot still forms. I prefer this method right now because the expected secondary structure is saved through the puzzle interface for later comparison to experimental results. The booster does not save the custom target structure, at least not yet.

Start by loading a known pseudoknot as explained above. I’ll use the tobacco virus again. Note how close the EternaFoldThreshknot predicts most of the structure correctly but does not find the second pair in the pseudoknot.

When I submitted this sequence, I got a warning message that one of my target structure pairs is not a standard pair. The last pair in the pseudoknot is GA if you look closely. I want to test if the pseudoknot still will form if I change that pair to AU or GC.

I changed base 52 to a U and EternaFoldThreshknot still does not predict the second pair in the pseudoknot. Submitting this sequence with the known structure (the custom target structure) will allow researchers to quickly see what I am testing. It’s also helpful to add a note in the description.

I changed base 108 to a C and now EternaFoldThreshknots predicts the entire pseudoknot! I will submit this design along with many other variations on the tobacco virus to help the algorithms learn more about predicting pseudoknot structure.

The known pseudoknot and changes you make don’t need to match what EternaFoldThreshknot is predicting. I only use this pseudoknot as a very clear example of structures to explore. If you see GA pairs in known pseudoknots, removing GA pairs (or adding GA pairs) is a useful structural characteristic to explore. Omei has proposed that these non-canonical pairs may be important for creating the chemical bonds needed to fold a pseudoknot. Testing (mutating) bases in the connecting loops also could provide useful insights.

1 Like

In the previous example, I used a PKB sequence that folded in the Eterna puzzle very similarly to the known structure. Now I’ll provide an example where the Eterna-predicted fold is different than the known structure to illustrate more clearly the role that the custom target structure can play in testing the limits of pseudoknots.

This is the known (experimentally-validated) structure of PKB 254. I put this structure in the puzzle using the Paste Target Structure feature.

Here is the Eterna prediction. It’s quite different.

I don’t want to test what Eterna is predicting because that is not how the sequence folds in the laboratory. The sequence folds into the structure I pasted into Target Mode. That is the structure I want to test.

This particular custom target structure is a little tough to navigate so I’ll switch into clash-free layout.

Once again, one of the hairpin loops has a non-canonical GA pair and the other hairpin loop has a non-canonical UU pair, which is another feature I’ve been noticing in a lot of pseudoknots. The bases are marked in the above screenshot. Or a player might want to explore changing pairs in the pseudoknot stems. There are lots of features to explore.

The point I’m highlighting is that being able to paste the known (experimentally-validated) structure in the puzzle gives us an easy way to thoughtfully explore the structure of known pseudoknots.