In this work we optimize a model that can be used to perform RNA secondary structure prediction using experimental information. Predictions are done using ViennaRNA. The model is trained by (a) applying to the prediction a bias that is a function of the experimental data and (b) iteratively improving the function so as to maximise agreement with experiment. Crucially, the paper reports a wide hyper-parameter scan that was used to train regularization hyper-parameters.
All the runs are done through a Jupyter Notebook that is included in the paper and in a linked GitHub repository. The hyper parameter scan requires the submission of a large number of jobs that can be performed in parallel. Our algorithms are implemented in Python using a number of standard packages (numpy/scipy/etc). In addition, it is necessary to compile the Python-enabled version of ViennaRNA.
The method is trained on the data that were available, but it is meant to be re-trainable as soon as new data are published. It would be great to be really sure that even someone else will be able to do it. In case we receive any feedback, we would be really happy to improve our Github repository so as to make the reproduction easier!
Resources
Associated event
