Functional Annotation of Proteomes Using Protein Language Models: A High-Throughput Implementation of the ProtTrans...

Area of Science:

• Computational biology
• Genomics
• Bioinformatics

Background:

• Accurate protein function prediction is vital for biological research but remains challenging, particularly for non-model organisms.
• Traditional sequence similarity methods often fail to annotate a significant number of proteins.
• Protein language models represent a significant advancement in improving functional annotation accuracy and comprehensiveness.

Purpose of the Study:

• To highlight the superior performance of the ProtTrans protein language model for per-protein annotation.
• To introduce FANTASIA (functional annotation based on embedding space similarity), a tool for large-scale annotation of uncharacterized proteomes using protein language models.
• To demonstrate the utility of FANTASIA through case studies in transcriptomics, model organisms, and non-model organisms.

Main Methods:

• Utilizing the ProtTrans protein language model for enhanced protein function prediction.
• Developing and implementing the FANTASIA tool for large-scale proteome annotation.
• Applying FANTASIA to three distinct biological case studies.

Main Results:

• The ProtTrans model demonstrates superior performance in per-protein annotation compared to other tools.
• FANTASIA effectively harnesses protein language model advances for efficient large-scale annotation.
• Case studies showcase FANTASIA's utility in transcriptomics, model organism gene function assignment, and non-model organism gene identification.

Conclusions:

• Protein language models offer a more precise approach to predicting protein function.
• FANTASIA provides a powerful and accessible tool for advancing functional annotation in diverse biological contexts.
• These advancements have the potential to significantly accelerate functional genomics and related fields.