Geometric deep learning of RNA structure

The methods are only described in the supplementary material. I think there are two problems. 1. Their formula for using spherical harmonics does not seem correct. There is no angle, only a dependence on a vector between a central atom and a neighbour. 2. They only consider the K nearest neighbours, but they do not seem to give a value for K. A competent referee should have noted these problems.

On top of that, this is not a prediction method. It is a score for judging structures generated by another method. In this case, the other method was always based on rosetta. This means the paper is describing a good scoring method for rosetta models. Its applicability to models from other sources is not clear.

Pretty big compared to the old RNA secondary structure prediction tools judging from the abstract. Wish someone put it on Sci-Hub… Maybe someone from HN will!