While I’m a fan of repeat bases in riboswitches, my look at the recent data tells me that too much or too little repeat, seemingly hurts cluster counts. So I decided to hit again with a new strategy.
I don’t think my old strategy necessarily is bad, since natural riboswitches have a high repeat ratio. However it might have reflected that many of natural switches that had a lot longer sequences than our labs and thereby also allow and need more extensive repeats.
Too little repeats, is also undesirable and will result in lower score.
Another option is that there might be something with our lab method that will not fully allow the same amount of repeat which can be found in natural riboswitches. Like some natural riboswitches sometimes had 4 or 5 G’s in a row. Something that is definitely a no go in our labs.
- Reward lab designs that keeps a repeat ratio between 30 - 40%
- Decrease reward for designs with 20 - 29%
- Decrease reward for designs with 41 - 50%
Penalize exponentially outside of that range.
This is for FMN/MS2 switches.
Turnoff labs will tend to be in the one end of the spectrum and Turnoff labs will tend to be in the other end. But this strategy is more like a catch them all and penalizing the worst.