I'm developing a learning program (SVM RNAi) for selecting good siRNA
targets. The goal is to select a few targets with the highest
probability of success for each gene. It is currently trained using
100 published siRNA sequences and 750 random sequences. The learning
program predicts >50% of siRNA designed by the rule-based algorithms
(e.g., Ambion, MIT, CSH) are not good candidates. This claim is
supported by the high cross-validation rate of the learning model.
With more experimentally demonstrated positive/negative siRNA
sequences, the performance of SVM RNAi will certainly improve. If you
have done siRNA experiments, could you share with me your positive and
especially negative results? Positive results can be found in
publications, negative results however are hard to find in public.
Your help will be greatly appreciated.
SVM RNAi can be found online at: