HIGHLIGHTS
SUMMARY
The authors also include more thorough experimentation on the most advanced architecture implemented (DProm)-a convolutional recurrent neural_network. Experiments include multiple training and testing datasets showcasing a comparison of different promoter annotations, sampling methods to overcome the imbalanced data, and the effect of output functions in neural_networks. While non-promoter sequences created from OPD helped models obtain slightly higher specificity than UD sequences when tested on hg38chr1, using non-promoter synthetic data to train models had disadvantages over non-synthetic data: all models trained on synthetic data had lower specificity than models trained on . . .
If you want to have access to all the content you need to log in!
Thanks :)
If you don't have an account, you can create one here.