Protein secondary structure prediction using a lightweight convolutional network and label distribution aware margin loss

作者：

Highlights：

• This is the first attempt to develop a lightweight convolutional network for PSSP.

• The channel shuffle is improved to fully achieve cross-group information exchange.

• The class imbalance of eight-class protein secondary structure data is considered.

• Our study demonstrates effectiveness of lightweight model for PSSP.

摘要

•This is the first attempt to develop a lightweight convolutional network for PSSP.•The channel shuffle is improved to fully achieve cross-group information exchange.•The class imbalance of eight-class protein secondary structure data is considered.•Our study demonstrates effectiveness of lightweight model for PSSP.

论文关键词：Protein secondary structure prediction,Imbalanced classification,Deep convolutional network,Lightweight network,Channel shuffle

论文评审过程：Received 30 December 2020, Revised 10 November 2021, Accepted 12 November 2021, Available online 1 December 2021, Version of Record 17 December 2021.

论文官网地址：https://doi.org/10.1016/j.knosys.2021.107771