m6A-Maize

Get Started

With the development of next generation sequencing (NGS), the research on RNA modification is also more and more intensive. Among that, N6-methyladenosine (m6A) modification as the most abundant existence, has significant relate to varieties of biological processes. At the same time, maize is most important food crop and cultivated throughout around the world, so study of m6A modification in maize has both economic and academic value. In this research, we proposed a weakly supervised learning model to predict the situation of m6A modification in maize. This model is based on meRIP-seq (low-resolution technology) data to predict the ranges which has possibility contain modification sites, and can well cope with the situation that species do not have single-base resolution data. In the training progress of model, the gated attention mechanism is used to better learn the weights of important instances and embed the score on bag-level. For the test dataset, our model perform favorable and the auROC achieves 0.8224 while auRPC achieves 0.8162. Furthermore, 10-fold cross-validation also proves the stability and accuracy of our model.