Are there any (perhaps prohibitively expensive) techniques that give high fidelity for transposons? This seems like a problem well suited to machine learning, given a dataset of paired high quality and low quality sequence data.