Fylo›ARCADIA›Graph
Hubs
Association·arcadia

Biological foundation models learn evolutionary rules

Claim that biological foundation models, trained on massive and evolutionarily diverse genome datasets, uncover emergent rules governing molecular evolution.

Confidence
90%
active

Evidence Quote

“By learning on such a massive scale, it's believed that the model will uncover an emergent set of rules governing molecular evolution”

Relationship

Biological foundation models learns Evolutionary nonindependence

Arguments

Biological foundation modelssubject
Evolutionary nonindependenceobject

Connections (6)

Petabase-scale sequence alignment increases viral discoveryAssociation
Reasoning on BFMs learning evolutionary rulesInferenceChain
Evolutionary nonindependence impacts biological foundation modelsInferenceChain
Limitations of phylogeny inference and utility of perplexity in BFMsInferenceChain
Co-evolutionary and phylogenetic relationshipsFactor
Evolutionary patterns explain COX1 sequence diversification and model limitationsInferenceChain

Evidence

“Reference on using evolutionary approaches to find organismal models for human biology.”

(2024). Leveraging evolution to identify novel organismal models of human biology doi:10.57844/ARCADIA-33B4-4DC5 ↗

“Evidence outlining new opportunities at the wild frontier for model organism research, supported by the cited references.”

(2025). Phylogenies and biological foundation models doi:10.57844/arcadia-znum-bm22 ↗

“Preprint on genome modeling and design using Evo 2 model”

Brixi G et al. (2025). Genome modeling and design across all domains of life with Evo 2 doi:10.1101/2025.02.18.638918 ↗

“Study using a language model to simulate extensive evolutionary time”

Hayes T et al. (2025). Simulating 500 million years of evolution with a language model doi:10.1126/science.ads0018 ↗

“Study on scaling models for protein generation and function understanding”

Bhatnagar A et al. (2025). Scaling unlocks broader generation and deeper functional understanding of proteins doi:10.1101/2025.04.15.649055 ↗

“Study showing protein language models learn phylogenetic relationships when trained on MSAs”

Lupo U et al. (2022). Protein language models trained on multiple sequence alignments learn phylogenetic relationships doi:10.1038/s41467-022-34032-y ↗

“Inquiry into whether protein language models capture phylogenetic information”

Tule S et al. (2024). Do protein language models learn phylogeny? doi:10.1093/bib/bbaf047 ↗

“Reference to the arcadiathemeR R package by Arcadia Science.”

(2024). arcadiathemeR link ↗

“Reference to the arcadia-pycolor Python package by Arcadia Science.”

(2024). arcadia-pycolor link ↗

“Reference to the textstat Python package.”

(2024). textstat link ↗

“"Evidence detailing NIH support for model organism research, associated with the cited reference "A Look at NIH Support for Model Organisms”

A Look at NIH Support for Model Organisms, Part Two

“Evidence summarizing publication trends in research involving model organisms, associated with the cited reference "Publication Trends in Model Organism Research".”

Publication Trends in Model Organism Research

“Evidence summarizing trends and insights on animal models used in preclinical gene therapy product studies, based on the cited reference "A review of animal models utilized in preclinical studies of approved gene therapy products: trends and insights".”

A review of animal models utilized in preclinical studies of approved gene therapy products: trends and insights

“Evidence summarizing mouse models of human disease from an evolutionary standpoint supported by the citation "Mouse Models of Human Disease: An Evolutionary Perspective".”

Mouse Models of Human Disease: An Evolutionary Perspective