NeurIPS-2022 PEER：A Comprehensive and Multi-Task Benchmark for Protein Sequence Understanding
Paper: PEER：A Comprehensive and Multi-Task Benchmark for Protein Sequence Understanding
PEER: A benchmark for evaluating protein sequence representations
Abstract
This work introduces a benchmark for evaluating protein sequence representations, covering protein function prediction, protein localization prediction, protein structure prediction, protein–protein interaction prediction, and protein–ligand interaction prediction. The authors also survey how different methods perform under multi-task learning; experiments indicate that large-scale pretrained protein language models achieve the strongest results.
Introduction
Inspired by ImageNet and GLUE, the authors seek to build a comprehensive protein benchmark with 17 biologically relevant tasks spanning diverse aspects of protein understanding. They evaluate CNNs, LSTMs, Transformers, and large-scale pretrained models.
Benchmark Tasks
Methods
Experiments
<HR align=left color=#987cb9 SIZE=1>