General Protein Pretraining or Domain-Specific Designs? Benchmarking Protein Modeling on Realistic Applications

TL;DR

This work introduces $\textbf{Protap}$, a comprehensive benchmark that systematically compares backbone architectures, pretraining strategies, and domain-specific models across diverse and realistic downstream protein applications.

Abstract

Recently, extensive deep learning architectures and pretraining strategies have been explored to support downstream protein applications. Additionally, domain-specific models incorporating biological knowledge have been developed to enhance performance in specialized tasks. In this work, we introduce $\textbf{Protap}$, a comprehensive benchmark that systematically compares backbone architectures, pretraining strategies, and domain-specific models across diverse and realistic downstream protein applications. Specifically, Protap covers five applications: three general tasks and two novel specialized tasks, i.e., enzyme-catalyzed protein cleavage site prediction and targeted protein degradation, which are industrially relevant yet missing from existing benchmarks. For each application, Protap compares various domain-specific models and general architectures under multiple pretraining settings. Our empirical studies imply that: (i) Though large-scale pretraining encoders achieve great results, they often underperform supervised encoders trained on small downstream training sets. (ii) Incorporating structural information during downstream fine-tuning can match or even outperform protein language models pretrained on large-scale sequence corpora. (iii) Domain-specific biological priors can enhance performance on specialized downstream tasks. Code and datasets are publicly available at https://github.com/Trust-App-AI-Lab/protap.

Figures (8)

• Figure 1: Overview of our benchmark Protap.
• Figure 2: Illustration of pretraining tasks in Protap.
• Figure 3: Illustration of downstream applications in Protap. (a) Biological process of the enzyme-catalyzed protein hydrolysis; (b) Biological process of targeted protein degradation by PROTACs, where PROTACs form a ternary complex with the target protein and E3 ligase, leading to target protein ubiquitination and degradation; (c) In protein-ligand interaction, the ligand binds to the protein pocket, blocking interactions with other molecules; (d) Protein function annotation prediction reveals the biological activities a protein participates in; (e) Mutation effect prediction optimizes protein properties or functions for protein engineering.
• Figure 4: Performance comparison between general architecture and domain-specific models across downstream applications.
• Figure 5: Scaling behavior in parameters and training data