Different types of switches

  • 4
  • Article
  • Updated 4 years ago
I have been trying to sort switches into types. What I hope to achieve is that it might help us in getting a better understanding of how to solve RNA switches in lab and how to make good switch lab candidates. This is a sum up of some of my Switch diary notes. Remember this is based on still rather few designs, so I'm telling what tendencies I see this far. Here are some of the main points as an introduction.

What makes a switch lab easy and hard:

- Higher amount of base pairs involved in the switch. Labs that have a low amount of actual switching bases are easier to solve.

- Switch labs that has a very big difference in numbers of base pairs between bound and unbound state, are harder to solve

- Full moving switches take higher % of of AU-pairs than partial moving ones. The partial moving ones take a higher GC amount than AU. (Labpuz02 exception to the latter)

- Double same turning pairs/dinucleotides are more common in the switching area (easier to split?)

- Weaker pairs in the switching area

- Higher amount of GU’s in unbound state (in switching area) than bound state. Unless Bound state has a higher amount of base pairs than unbound, then there is a tendency for reversal. (Read more here) Jnicol already noticed that one state took more than the other. (Link)

- Nando and Brourd considers non-canonical bonds as important in the switching area. I think they can very well be right about that.

- If a whole hairpin in the unbound state is involved in the switching area, it is a bad idea of boosting the end loop too effective. Making it less negative however, can help the switch on the way. It has to break so it can get moving.

- Weakening of multi loop in unbound shape to make Bound state form (If the multi loop is involved in the switching area)

-There also seems to be a tendency for the labs with more than one internal loop, to be harder Higher amount of internal loops in the switching area (no matter if in bound/unbound state) seems to make the labs harder.

-Generally the bound state got 1 more stem than the bound shape. Which is quite logical, since the bound state contains the aptamer loop, which takes an extra stem. Some of the lower scoring designs have a difference that is bigger, however FMN aptamer 2.0 is an exception

SWITCHING AREA - FULL OR PARTIAL MOVE

There are different types of switches, somewhere only part of the two shapes move in relation to each other and switches where both shapes move in relation to each other. It wouldn’t surprise me if the different types of RNA switches has a different nature and will need a different approach.



I split the switch labs into types and noted the score % of the top scorer.

Full move
Switch test 1 (82%)
Lines (45%)
Nincompoop (78%)
My screw-up corrected (99%)
Water snake redux (88%)
McSwitch (85%)
Day 2 (86%)
LS2 (88%)
Will it bind? (96%)
FMN Switch 2.0 (92%) (mixed pattern)
FMN Aptamer 2.0 (95%) (mixed pattern)
Simple RNA Switch (94%)
(Higher % of AU than GC)

Partial move (more than barcode hairpin doesn’t move)
Chirality (83)
Bound state has a zigzag by steveclark (73)
Lay down and jump (90%)
Alive and kicking (89%)
Anchor (81%)
Phase two (84%)
Stratospheric (93%)
Hair trigger (95%)
Bistable (88%)
LabPuz02 (96%) (Seems to be an exception)
(Higher % of GC than AU)

Small move
Top notch (100%)

Those which change fully, has a tendency to want to have more AU than GC. Which in itself is quite logical, as the AU is weaker than GC, and therefore is easier to break, which is necessary when all the base stations need to move spot relative to each other. Yet it is quite nice to know in advance that because this design change completely, so it must have fewer than AU GC. The only partially changing generally want more GC pairs than AU.

It seems the switches that only moves a little are fairly rare so far. So far it seems like the small move ones and second the partial moving ones might have an easier time getting a higher score than the full moving ones. And that highest scoring one in the full move category. (My screw-up) is a quite small puzzle and similar the long stringed (Will it bind?) is a quite different kind of puzzle.

POSITION OF LOCKED NUCLEOTIDES

Again other types might be special type of switches. Eg like this one where locked bases moves from aptamer (second state) to loop area (first state). This might behave differently to designs where the locked nucleotides mainly end up in stem area. What makes a switch lab solve possible will of cause be a combo of many things. The placement of the locked nucleotides, seems to play a role too. Especially if the A’s and U’s of the aptamer gets locked up in a short stem in the switching area, as those can only make weak pairs. Read more about that here. I think there is much more to be found out about this.

Aptamer becomes stem
Aptamer locked nucleotides become loops
Aptamer locked nucleotides becomes both stem and loop.
Aptamer locked nucleotides becomes multi loop and stem

BIG DIFFERENCES IN NUMBERS OF BASE PAIRS BETWEEN THE STATES

The percentage is for the highest score in that lab. The numbers afterwards are the number for the differences in base pairs among the unbound and the bound shape. (Hairpin barcode not counted in.) I decided to add a minus in front of the labs that had more pairs in the unbound shape. And sorted them after score.

Top Notch (100%) -1
My screw-up corrected (99%) -5
LabPuz02 (96%) -2
Will it bind? (96%) 1
FMN Aptamer 2.0 (95%) 3
Hair trigger (95%) 0
Simple RNA Switch (94%) 1
Stratospheric (93%) 0
FMN Switch 2.0 (92%) -4
Lay down and jump (90%) 2
Alive and kicking (89%) 3
Bistable (88%) 0
LS2 (88%) 1
Water snake redux (88%) -3
Day 2 (86%) 1
McSwitch (85%) -3
Phase II (84%) -1
Chirality (83%) 0
Switch test 1 (82%) 1
Anchor (81%) -6
Nincompoop (78%) 3
Bound state has a zigzag by steveclark (73%) 1
Lines (45%) 7

I see that Lines and Anchor, probably have too big a difference in base pairs, between the two states.

NUMBER OF BASE PAIRS IN THE SWITCH AND HARDNESS

I added in parenthesis is the number of base pairs in bound and unbound state. ( ) = (bound-unbound)

Top Notch (100%) -1 (17-18)
My screw-up corrected (99%) -5 (4-9)
LabPuz02 (96%) -2 (13-15)
Will it bind? (96%) 1 (24-23)
FMN Aptamer 2.0 (95%) 3 (10-7)
Hair trigger (95%) 0 (16-16)
Simple RNA Switch (94%) 1 (7-6)
Stratospheric (93%) 0 (16-16)
FMN Switch 2.0 (92%) -4 (12-16)
Lay down and jump (90%) 2 (18-16)
Alive and kicking (89%) 3 (20-17)
Bistable (88%) 0 (16-16)
LS2 (88%) 1 (14-13)
Water snake redux (88%) -3 (16-19)
Day 2 (86%) 1 (17-16)
McSwitch (85%) -3 (16-19)
Phase II (84%) -1 (16-15) [10-11]
Chirality (83%) 0 (14-14)
Switch test 1 (82%) 1 (16-17)
Anchor (81%) -6 (13-19)
Nincompoop (78%) 3 (14-11)
Bound state has a zigzag by steveclark (73%) 1 (22-23)
Lines (45%) 7 (20-13)

Here a pattern does emerge. Top notch stick out from it. But the switches with a higher number of base pairs pr puzzle, seems to have a harder time getting a high score. However if you do remember, Top notch is the only one and rare lab, where only a small portion of the base pairs are actually switching. So for now it looks like switch ability is somehow coupled with number of switching base pairs.

NUMBER OF ACTUAL SWITCHING BASE PAIRS AND HARDNESS

I decided to look at the numbers of actual switching base pairs. Labs sorted after top scorer. Number for difference in base pairs in ( ). Number for difference in actual switching base pairs in [ ].

Top Notch (100%) -1 (17-18) [5-6]
My screw-up corrected (99%) -5 (4-9) [4-9]
LabPuz02 (96%) -2 (13-15) [4-6]
*Will it bind? (96%) 1 (23-24) [23-24]
FMN Aptamer 2.0 (95%) 3 (10-7) [10-7]
Hair trigger (95%) 0 (16-16) [8-8]
Simple RNA Switch (94%) 1 (7-6) [7-6]
Stratospheric (93%) 0 (16-16) [4-4]
FMN Switch 2.0 (92%) -4 (12-16) [12-16]
Lay down and jump (90%) 2 (18-16) [15-12]
Alive and kicking (89%) 3 (20-17) [10-7]
Bistable (88%) 0 (16-16) [4-4]
LS2 (88%) 1 (14-13) [14-13]
Water snake redux (88%) -3 (16-19) [16-19]
Day 2 (86%) 1 (17-16) [17-16]
McSwitch (85%) -3 (16-19) [16-19]
Phase II (84%) -1 (16-15) [10-11]
Chirality (83%) 0 (14-14) [8-8]
Switch test 1 (82%) 1 (16-17)
Anchor (81%) -6 (13-19) [6-12]
Nincompoop (78%) 3 (14-11) [12-9]
Bound state has a zigzag by steveclark (73%) 1 (22-23) [13-14]
Lines (45%) 7 (20-13) [20-13]

It looks like the labs with the so far highest scores, have a relatively low amount of actual switching bases. It seems very logical that it is easier to get a small amount of base pairs to switch fully than to do the same with a huge amount of base pairs.

The long stringed design Will it bind, sticks out from the picture. So is it easier to get a stem to move along itself? It could really look like it. I have also been wondering about the pattern I have been seeing in the lab solves. Repetitive patterns that I would normally have deemed bad for the stems sticking ability seemed to have thrived. Like lots of AU’s totally not alternating, something that sometimes do causes slide. (Which just happen to be very useful here.) The highest scoring designs overall has this pattern, except Starry’s design that only has it to some degree. They are stitched up with a few GC-pairs and some GU’s pairs to help the switch happen.

This pattern (GUG...) if in excess usually causes trouble in static labs.

Starrys winner:


I think I see a tendency for designs having more difference in numbers of base pairs between bound and unbound state, also have a lower chances of getting a high score. With designs having a small switching area, scoring high anyway. Like My screw up corrected. Labs with low scores like Lines have both a high difference in base pairs (7) between bound and unbound state and a big switching area (20-13). I think both factors, having a big switching area + having many base pair differences between bound and unbound state, adds to the difficulty of solving a switch lab.

Unbound state has most pairs
Top Notch (100%) 1
My screw-up corrected (99%) 5
LabPuz02 (96%) 2
FMN Switch 2.0 (92%) 4
Water snake redux (88%) 3
McSwitch (85%) 3
Phase II (84%) 1
Anchor (81%) 6

Bound state has most pairs
Will it bind? (96%) 1
Simple RNA Switch (94%) 1
FMN Aptamer 2.0 (95%) 3
Lay down and jump (90%) 2
Alive and kicking (89%) 3
LS2 (88%) 1
Day 2 (86%) 1
Switch test 1 (82%) 1
Nincompoop (78%) 3
Bound state has a zigzag by steveclark (73%) 1
Lines (45%) 7

Equal amount of pairs in both states
Hair trigger (95%) 0
Stratospheric (93%) 0
Bistable (88%) 0
Chirality (83%) 0

HAIRPIN IN SWITCHING AREA

Another thing I think is a bad idea, for the stem in the switching area, is to boost the end loop so it gets more stable. The few of the high scorers that did that and actually did make a switch in the right direction, had the loop area weakened. Either like below with a weak closing pair/backing pair, or by another boost in the loop, that weakened the loop. So the more towards negative the tetraloop becomes, the more stable it gets and the less likely it is that the switch is happening. Read more about it here.

MULTILOOPS IN THE SWITCHING AREA

Making bound state form by weakening of multiloop in unbound shape

NUMBER OF GU-PAIRS - BOUND/UNBOUND

I have checked the distribution of GU’s in the absolute top scorers in the different lab, trying to note if there was a pattern.

There is a tendency for the shapes that does overall worse when it comes to score, the top scorers have a more reversed pattern for GU distribution, where the labs that score has highest has a tendency of more GU’s in the unbound state than the bound state.

- There is a tendency for the designs with Equal amount of GU in both states, to be designs that has a very similar amount of switching base pairs in each state.

- The designs with more GU in bound state are among the designs that has a high amount of base pairs in bound state compared to unbound. (Something that in itself makes the switch harder.) So it is actually quite logical that the balance of GU’s in in bound versus unbound, shifts toward bound, when bound state has the most base pairs. It is the state with most base pairs compared to the other, that need some pushing, to be made weaker. This Jnicol has also noticed for the puzzles earlier. (See this (link)

I think switch labs will behave the same way in regards to this as the puzzles.

Also take note that each of the switching arms have a GU pair in this design. The second highest scoring design in this lab. Something similar goes on in the lab McSwitch. It is a full move lab and has 3 longer stems in the unbound state. The highest scoring designs have GU’s in more than one of the stems.



Likewise take note that Brourd has put two AA pairs in each of the stems in the Bound state (that has a lot more basepairs than the unbound state. This design has the highest score in this, the so far hardest of the switch labs.



NUMBER OF STEMS IN THE TWO STATES

There is a tendency for more stems in the lower scoring designs. (When looking at the switching area itself - not counting all stems). That also fits with that there is an overall higher amount of base pairs among the general lower scoring labs.

Overall the bound state got 1 more stem than the bound shape. Which is quite logical, since the bound state contains the aptamer loop, which takes an extra stem. Some of the lower scoring designs have a higher difference in the number of stems between bound and unbound state. However FMN aptamer 2.0 sticks out from that picture. As it has 1 stem in unbound and ends up with 4 in bound. If there is more stems (in the switching area) in the unbound states, compared to the bound state, that might also mean trouble, as there was a few of those among the lower scoring labs. But more numbers will be needed to truly to say something about that.

NUMBER OF INTERNAL LOOPS

Among the highest scoring labs, there is fewer internal loops, in unbound state. Top notch is the exception, but this one is easy for other reasons, given it has a really small switching area.The lower scoring labs have a higher amount of internal loops in unbound state and bound state too.

PRESENCE OF MULTI LOOP OR NOT

Designs with multi loops in either bound state or unbound or both, overall did well, if not taking into account designs as Anchor, that had a big difference in base pairs between the two states (6) or Phase II that had a huge amount of actual switching pairs ( 15-16). It could look like designs with multi loops have a slight advantage.
Photo of Eli Fisker

Eli Fisker

  • 2183 Posts
  • 469 Reply Likes

Posted 5 years ago

  • 4
Photo of nando

nando, Player Developer

  • 388 Posts
  • 71 Reply Likes
I will finish to "digest" this later, but it already inspires me a few reactions...

So, it seems I'm not the only one who misses our messy riboswitches ;) I have to admit that, contrary to what I had imagined a few months ago when Cloud Lab was launched, mono-state lab puzzles are still teaching us quite a lot. I do regret though, that we aren't anymore given the chance of testing certain ideas on switching labs.

Impressive work, Eli, as usual :)
A remark though: I have an intuition that static features will not be sufficient to give us very effective design guidelines. I believe that we will need to get a better grip of thermodynamic and kinetic features of the sequences. For the moment, we may be lacking the proper computational tools for such kind of analysis though (read: dotplots and meltplots can't get us very far in the context of switches)
Photo of Eli Fisker

Eli Fisker

  • 2183 Posts
  • 469 Reply Likes
Hehe, thx Nando.

I will look forward to hear you add your thoughts on the switches.

There is so much more to be found out, when it comes to the switches - and static puzzles as well. I'm happy this being an inspiration.

Though I think dot plot and melt plot can still be of help, they are far from as strong guiding tools, as for the static puzzles.

For every player, feel free to add your piece to solve the mystery of the switch puzzle.
Photo of nando

nando, Player Developer

  • 388 Posts
  • 71 Reply Likes
For those who missed the dev chat:


Nando: given one of Eli's recent forum post, it seems natural to ask: how long will we wait for the "Return of The Switches" (IMAX, err, in Cloud lab) ? [3:27 PM]
jeehyung: @Nando we have a few renovations to be made on the cloud lab. That might take 2-3 weeks. I think switch will come after that at the earliest. [3:28 PM]
jeehyung: We do want to visit the theophylline lab again soon so hopefully it won't be too long. [3:28 PM]


It seems we won't have to wait too long :)
Photo of Eli Fisker

Eli Fisker

  • 2183 Posts
  • 469 Reply Likes
I was looking at one of the old switch labs while preparing for the current round of mimics. I mentioned the following in the post above:

- If a whole hairpin in the unbound state is involved in the switching area, it is a bad idea of boosting the end loop too effective. Making it less negative however, can help the switch on the way. It has to break so it can get moving.

Since there is a small hairpin involved in the switching area I was thinking about what would be the best closing base pair for the hairpin loop, if I made it a GC.

My first thought was to use the opposite pattern to what I think is really the most effective (if not overused) for stabilizing a hairpin loop. (Besides the G boost) As to not make it too stable, so it can break for a switch.

My favorite


My second favorite


Also notice that there is a lower energy in the loop to the left.

I did a query in the LabDataMiner. I kicked classic eterna results out as I knew that all A’s tetraloops were favored way too much in that dataset.

Favorite


Second favorite


My favorite pattern did best

Then I decided to look up the past switch lab, which I knew had a breaking hairpin.



And interestingly enough the switch designs which score highest seems in favor of the slightly less stable closing.



Highest scoring designs have a higher prevalence of G’s of this closing base spot and the reverse pattern get more prevalent for the lower scoring designs. And that is despite that most haven’t even boosted the tetraloop.



It will be interesting to see if this tendency continues when we get more data for switches.