Design rules for DNA linkers

Question

I want to use double stranded DNA linkers to physically bind two "things" together, by grafting ssDNA on each one of them and using DNA hybridization as the locking mechanism.

I do not expect the nature of the two things to be important for this question, but if it is, let's say I want to link a solid surface and a protein.

Let's say that my only constraint is that I want this crosslinking to be stable at 37°C.

There is a gazillion ways I could design the sequences, and pretty much anything above 20bp without massive self complementarities should work, in theory. But I don't like to pick bases randomly. So my question is this: is there anything I can look for, except for the two constraints above, to design a good linker sequence?

For example, if the melting temperature is 50°C vs. 65°C, I would expect the structure to be stable (infinitesimal dissociation constants in both cases). Would there be any reason to choose one over the other, except for maybe oligo synthesis price? For a given melting temerature, is there any difference between a long AT rich sequence and a short GC rich sequence?

Answer 1

You might want to pick something that has little or no homology with anything that exists in any organism in the databases, just in case.

Gut feeling says to go for balanced oligos in the middle between AT-rich and GC-rich, and not have repeats or homopolymers. If you go for AT-rich, then your complementary will be GC-rich, and vice versa. I'm pretty sure there would then be a differential in the mechanical properties of the two oligos - maybe this would be something useful to have for other reasons though?

It seems like most of the design principles might be constraints (do nots) rather then goals (dos), so you could do worse than have a generator churn out random sequences, and score each one against the various constraints. Random seems OK, if you are then checking each randomly chosen pair against some strong rules. Then pick ten or so pairs that all score well, get them synthesised and see whether any work...