Is it possible for me to recreate the search you did with the web interface at https://blast.ncbi.nlm.nih.gov/Blast.cgi?PAGE_TYPE=BlastSearch? The one I did for my GPB5 proposal (see above) found no exact match in the human genome other than the GPB5 gene itself. I wouldn’t be in the slightest surprised if I didn’t perform the best search. But if I could reproduce yours, I could better understand what we might be giving up in considering other choices.
Couldn’t you just change the sequence of the MS2 hairpin to be complementary to any oligonucleotide sequence of choice?
@Omei: I can’t agree with your assertion that it is useless to work with the actual TB-related sequences. A functioning device is necessarily dependent of both inputs and outputs. Working on the inputs-device dependency seems very useful, to the contrary, while avoiding inputs-output dependence would allow us to replace the output by anything we want.
Also, what Brourd means to say, is that you don’t even need to select TB-oligos to match the standard MS2 sequence. Instead, if you insist on developing methods based on sequence complementarity and inputs-output dependence, then you can simply use one of the alternative MS2 sequences offered by the MS2 painting tool. One of these 25 possibilities is bound to give you what you’re looking for.
Full length versions of
[A]/[C] http://nando.eternadev.org/web/puzzle/3398879/
[B]/[C] http://nando.eternadev.org/web/puzzle/3398881/
1 Like
Please ignore my previous posts. They were based on erroneous assumptions.
@omei What you did was exactly right. I just pasted in the full 50 nt probe sequence instead. You won’t find other exact matches, but you will find other genes that match up to 20 nts of the 50.
@Nando: I certainly wouldn’t agree with an assertion that it was useless to work with the actual TB–related sequences, either. I’m not sure how you read it that way, but it doesn’t really matter – I’ll disavow it. 
I’ve been thinking about why, intuitively, it seems good to start by trying for the most sensitive switch (i.e. the largest fold change) we can, even though that’s certainly not the only relevant criteria for a final product. I’ve decided it’s motivated by my own development experience in a totally different domain (writing embedded software for novel hardware, where both are being developed simultaneously.)
Imagine a scenario where you are part of a project developing a novel radio technology that has a chance to be the standard for the next generation of cell phones. Everything about the technology is new, so if things aren’t working well at all, it could be a hardware problem, a software problem, bad theory, operator error, … the list goes on and on. So you start off testing in a very controlled environment – in the lab, with the strongest signal and the best shielding you can get. Only after you get that working reasonable well do you start testing more real-world scenarios.
I’ve been viewing the development of a TB diagnostic based on Eterna generated switches in a similar light. I have only the vaguest of ideas about what it will take to translate Johan’s experimental results into a practical real life diagnostic. But I have to believe that the path will be full of potholes, and nobody really knows where they will turn up. It seems like there is real value in trying to start the downstream researchers off with the strongest “signal” we can give them. Undoubtedly, as they and we learn more, it will become clearer what we should be optimizing for, and we will modify our designs accordingly.
So FWIW, that’s my rationale.
I want to add a comment on this bit:
“I don’t think that the oligos we’ve been using until now were particularly selected for their compatibility with the MS2 signal…”
What I found when it came to the FMN aptamer and the MS2 was that they were actually a perfect match with each other, which I by the way was actually rather surprised about, since one come from bacteria and the other from virus and likely never in any way have been interacting with each other to my knowledge.
I have written a bit about it earlier. Search for this section The FMN piece inside the MS2 hairpin in the piece.
Good idea, Nando.
Assuming it will be quick, could you make yet another version where the oligo bases aren’t locked? (Or, even better, is there some back door to unlock specific ones?) The rationale is that if it comes to a question of how susceptible a design is to folding by a specific non-TB RNA sequence, we could see what NUPACK thinks.
Plus I agree here:
“It seems like there is real value in trying to start the downstream researchers off with the strongest “signal” we can give them. Undoubtedly, as they and we learn more, it will become clearer what we should be optimizing for, and we will modify our designs accordingly.”
Its like when we have the main rules in place for what works - then we can easier go for the finer details of what we need to learn also. Something which isn’t as easily seen when everything is one big mess.
@Nando, I actually like your new A/C puzzle with full length gene far better than the first version of the A/C puzzle.
The original A sequence is rather weak - really it is a whole long bunch of A’s with a few G’s on plus a strong magnet stretch. The new full puzzle, allowed me to pick out a better start sequence than the original one, fit it to the MS2, plus I ended making a far simpler solve.
@Brourd, @Nando: Do you know of any systematic experiment anyone has done on how the alternate MS2 hairpin sequences affects the KD values? If not, we ought to make sure someone does that this round.
Actually, Brourd, it looks like you did just that in the first A/B round. Have you summarized your findings?
@Omei: the Greenleaf lab did a quite intensive series of experiments on the MS2:
http://www.nature.com/nbt/journal/v32/n6/full/nbt.2880.html
The information you’re looking for is in Table 2 & 3 (Supplementary Informations, at the bottom of that page)
If you don’t have access to the article itself, the Greenleaf lab has it on its webpage: http://greenleaf.stanford.edu/portfolio_details_buenrostro_2014_nature_biotechnology.html
Oligos can’t be unlocked, no more than the FMN or TEP molecules can be changed. They’re considered as being part of the environment for the experiment, just like other factors like salt concentration and temperature.
The only place where this manipulation could be considered, is in the puzzlemaker, but upgrading it to support oligos is a monumental task that I don’t expect to be accomplished any time soon…
Thanks, Nando!
In case anyone else might have the same trouble I had in finding where to get the article text, it’s the magnifying glass that might be mistaken for one of a set of social media icons.
My take on these new puzzles have to be aiming for a stretch of the single base region, be it tail dangle or loop region. This is enough to get hold of the paired up sequences. I have been avoiding the strongly paired up regions in the hairpins where there is double same turning GC pairs.
Thx, Omei! I had.
I think it is important to mention that using 50-nts models might cause troubles. You’ve seen the performance hit on a 4-states puzzles with 4 oligo instances. Adding 4x30 = 120 bases on top of that is most likely going to make the solving intolerably slow, no matter what trick I’ll come up with…
Hi Nando! Point taken. The main thing I’m happy about is that I could pick a better working sequence for the puzzle to make a simpler solve. After that I didn’t need the rest of the bases.
Far too tired, Omei. However, the general consensus based on the data is that mutations to the helix preceding the MS2 hairpin result in Kd values similar to the wildtype sequence as long as the helix is of sufficient length and is not entirely made of G-C base pairs pr consists of a 4x4 internal loop. However, these experimental sequences had kD values less than 20 nM, indicative of a fairly decent rate of association between the protein and MS2 hairpin.
Therefore, you can design the MS2 hairpin however you want to design it.