This issue is exactly one of my research projects. So just give my 2 cent. 1.the efficacy of siRNA vary dramatically against the same target gene. By far, there is NO clear rule of selection. 2.ALL software can achieve no more than 70% of accuracy of prediction, no matter where you find it or whatever they claim themselves could achieve. 3.There probably endless factors which could affect the efficacy of siRNA. The real efficacy is combination of all, though some factors do contribute more other
from the work of my hand: genarally speaking, select blindly you will get 1 out of 2 working, 1 out of 5 will working really good ( >80% knock down) But i also know a guy select 19 sites, and none of them work.. hehe, poor guy...