StoPred: Accurate Stoichiometry Prediction for Protein Complexes Using Protein Language Models and Graph Attention

Area of Science:

• Computational Biology
• Structural Biology
• Bioinformatics

Background:

• Protein complexes are crucial for biological functions, but determining their subunit stoichiometry is experimentally challenging.
• Existing computational methods for stoichiometry prediction have limitations, especially for proteins lacking close homologs or known assembly states.
• Current protein language model (pLM) approaches predict homo-oligomer stoichiometry but fail with hetero-oligomeric complexes and do not fully model inter-subunit relationships.

Purpose of the Study:

• To develop a novel computational method, StoPred, for accurate prediction of protein complex stoichiometry.
• To address the limitations of existing methods by enabling prediction for both homomeric and heteromeric complexes without requiring templates or predefined compositions.
• To leverage advancements in pLMs and graph attention networks for modeling subunit interactions.

Main Methods:

• Integrated protein language model (pLM) embeddings with a graph attention network (GAT).
• Modeled subunit-level interactions within protein complexes.
• Inferred stoichiometry directly from sequence or structure features for both homo- and hetero-oligomers.

Main Results:

• StoPred demonstrated improved accuracy and efficiency compared to deep learning-based and template-based methods.
• Achieved up to 16% higher top-1 accuracy for homomeric and 41% higher for heteromeric complexes on a held-out test dataset.
• StoPred is the first deep learning method capable of accurately predicting hetero-oligomeric complex stoichiometry.

Conclusions:

• StoPred offers a significant advancement in predicting protein complex stoichiometry, particularly for hetero-oligomeric assemblies.
• The method's ability to predict stoichiometry from sequence or structure without prior knowledge enhances its applicability.
• StoPred provides a powerful new tool for computational biology and structural biology research.