PreRBP-TL:

prediction of species-specific RNA-binding proteins based on transfer learning

| Home | Server |



Introduction

PreRBP-TL is a new sequence-based computational predictor for identifying species-specific RNA-binding proteins (RBPs). In which hybrid convolutional neural networks (CNNs) and transfer learning were employed. The weights of prediction model were initialized by pre-training with the large general RBP dataset, and then fine-tuned with the small species-specific RPB dataset by using transfer learning. The experimental results show that the PreRBP-TL achieves obviously higher performance for identifying the species-specific RBPs from six different organisms, outperforming eight state-of-the-art computational methods, which is an important complement to existing methods. The network architecture and framework of PreRBP-TL are shown in Fig. 1. For more information please refer to our paper.

The network architecture and framework of PreRBP-TL
Fig 1. The architecture of hybrid convolutional neural networks (A) and the overall framework of PreRBP-TL (B). The hybrid CNNs consisting of the traditional CNNs and the motif-based CNNs are used to excavate known RBP-related patterns and potential RBP-related patterns, respectively. The fully-connected feed forward network is used to provide a nonlinear transformation for the learned hidden features so as to learn the correlation between RBPs and various patterns. Only the weights in the yellow boxes are trained in the processes of initially training and fine-tuning.

Materials

The structural motifs used in this study: motifs.txt

Pfam RNA-binding domains used in this study: RBDs.txt

All the datasets used in this study: Datasets.zip

Acknowledgments

We acknowledge with thanks the following softwares used as a part of this server:

(1) PSI-BLAST - Generation of the position specific scoring matrices (PSSMs);

(2) NRDB90 - The non-redundant database for usage of PSI-BLAST to generate PSSMs;

(3) Biopython - Protein data processing;

(4) Keras - Construction of the prediction model;

(5) Tensorflow - Backend of keras for calculation.

References

Upon the usage of this server the users are requested to use the following citation:

Jun Zhang, Ke Yan, Qingcai Chen, Bin Liu*. PreRBP-TL: prediction of species-specific RNA-binding proteins based on transfer learning. Bioinformatics,2022,38(8):2135-2143.