Translating a GO Term List to Human Readable Function Description Using GO2Sum

Swagarika J Giri; Udayan Pandey; Joon Hong Park; Daisuke Kihara

doi:10.1007/978-1-0716-4623-6_5

Translating a GO Term List to Human Readable Function Description Using GO2Sum

Methods Mol Biol. 2025:2941:85-99. doi: 10.1007/978-1-0716-4623-6_5.

Authors

Swagarika J Giri¹, Udayan Pandey¹, Joon Hong Park¹, Daisuke Kihara^{2

3}

Affiliations

¹ Department of Computer Science, Purdue University, West Lafayette, IN, USA.
² Department of Computer Science, Purdue University, West Lafayette, IN, USA. dkihara@purdue.edu.
³ Department of Biological Sciences, Purdue University, West Lafayette, IN, USA. dkihara@purdue.edu.

PMID: 40601252
DOI: 10.1007/978-1-0716-4623-6_5

Abstract

Understanding the functions of proteins is one of the most important challenges in modern biology. Typically, protein function prediction methods generate a list of gene ontology (GO) terms, sometimes consisting of 50-100 functional terms. While GO serves the purpose of standardizing terms, interpreting a long list of GO terms is difficult for biologists. To address this challenge, we developed Gene Ontology terms Summarizer (GO2Sum), a language model-based summarizer that takes a list of GO terms as input and converts them into a concise, free-text summary describing the protein's function, subunit structure, and pathway information. GO2Sum was fine-tuned on GO term assignments and free-text function descriptions from UniProt entries. We built a Web server of GO2Sum, which offers an easy use of GO2Sum for biology users.

Keywords: Function prediction; GO2Sum; Gene ontology; LLM; Large language model; Protein function; Summarization.

MeSH terms

Computational Biology* / methods
Databases, Protein
Gene Ontology*
Humans
Proteins* / chemistry
Proteins* / genetics
Proteins* / metabolism
Software*

Substances

Proteins