LongevityLLM: A Function-Driven AI Agent for End-to-End Protein and Aging Research

Recent advances in large language models (LLMs) have unlocked new possibilities for scientific discovery, yet most remain limited to text summarization or hallucination-prone dialogue. Here, we present LongevityLLM—a function-driven AI agent engineered to execute real, reproducible analyses in structural bioinformatics, comparative genomics, and aging biology. Unlike conventional chatbots, LongevityLLM maps natural language queries to deterministic bioinformatics pipelines, producing structured outputs (FASTA, PDB, XLSX, phylogenetic trees, aging clock reports) while grounding all responses in empirical data. The system retrieves and summarizes scientific information from peer-reviewed literature (via Europe PMC) and biological databases (e.g., UniProt). It integrates five major epigenetic clocks—Horvath, Hannum, PhenoAge, Brunet, and Wyss-Coray—as well as AlphaFold2-based structural mutation impact prediction, cross-species ortholog retrieval with phylogenetic analysis, and curated mammalian life-history traits from the AnAge database and incorporates a time-calibrated mammalian phylogeny and the AROCM (Average Rate of Change in Methylation) metric—a cross-species epigenetic biomarker of aging derived from conserved CpG sites. Built on open-source tools and designed for full auditability, LongevityLLM enables researchers to explore questions such as “How is IFI27 implicated across different aging clock models?” or “What is the structural effect of the IL17A-E100K mutation?” through a single natural language query, without compromising scientific rigor. We release LongevityLLM as an open framework to accelerate hypothesis generation, education, and collaborative geroscience.