Fig. 1: The pipeline of DRfold for deep learning-based RNA structure prediction by combining end-to-end model and geometry potentials. | Nature Communications

Fig. 1: The pipeline of DRfold for deep learning-based RNA structure prediction by combining end-to-end model and geometry potentials.

From: Integrating end-to-end learning with deep geometrical potentials for ab initio RNA structure prediction

Fig. 1

A DRfold pipeline for sequence-based RNA structure prediction, where \({D}_{s}\) and \({D}_{z}\) are hidden dimension sizes of sequence and pair features, respectively, and L is the length of the query sequence. B–E Details of embedding layer, RNA transformer block, and structural and geometry modules, respectively. F Reduced representation of nucleotide residues by a 3-bead model (C4’, P, glycosidic N) in DRfold. G Illustration of the frame aligned point error (FAPE). H Prediction terms of inter-nucleotide geometry.

Back to article page