207: Semantic Design of de novo Genes with Evo

Semantic Design of de novo Genes with Evo

Music:
Enjoy the music based on this article at the end of the episode.

DOI:
10.1038/s41586-025-09749-7

License:
This episode is based on an open-access article published under the Creative Commons Attribution 4.0 International License (CC BY 4.0) – https://creativecommons.org/licenses/by/4.0/

Support:
Base by Base – Stripe donations: https://donate.stripe.com/7sY4gz71B2sN3RWac5gEg00

Official website https://basebybase.com

On PaperCast Base by Base you’ll discover the latest in genomics, functional genomics, structural genomics, and proteomics.

Episode link: https://basebybase.com/episodes/semantic-design-of-de-novo-genes-with-evo

️ Episode:
207: Semantic Design of de novo Genes with Evo

️ Season:
1

Article title:
Semantic design of functional de novo genes from a genomic language model

Journal:
Nature

QC:
This episode was checked against the original article PDF and publication metadata for the episode release published on 2025-11-24.

QC Scope:
- article metadata and core scientific claims from the narration
- excludes analogies, intro/outro, and music
- transcript coverage: Audited transcript sections describing Evo 1.5 semantic design, contextual autocompletion, toxin–antitoxin designs (T2TA/T3TA), anti-CRISPR designs, and SynGenome, plus limitations and licensing statements.
- transcript topics: Semantic design and genomic context; In-context design and autoregression; Toxin–antitoxin (T2TA) design and validation; Toxin–antitoxin (T3TA) design and validation; Anti-CRISPR (Acr) design and validation; SynGenome database and analyses

QC Summary:
- factual score: 10/10
- metadata score: 10/10
- supported core claims: 6
- claims flagged for review: 0
- metadata checks passed: 4
- metadata issues found: 0

Metadata Audited:
- article_doi
- article_title
- article_journal
- license

Factual Items Audited:
- Semantic design uses genomic context to enable function-guided design of de novo genes (Evo 1.5).
- Autocomplete test: Evo 1.5 achieved 85% amino acid recovery for rpoS with 30% input.
- Toxin–antitoxin (T2TA) design yielded functional toxins and antitoxins; EvoRelE1 toxin; EvoAT1–4 antitoxins; EvoAT2 and EvoAT4 show multitoxin neutralization; EvoT3TA design produc
- Anti-CRISPR (Acr) designs produced functional Acrs; 17% of tested Acrs exhibited activity against SpCas9; EvoAcr1–5 validated.
- SynGenome database contains over 120 billion base pairs of AI-generated DNA; ~3.7 million predicted protein structures; Pfam-domain frequencies in SynGenome closely mirror natural
- Limitations include autoregressive generation risks (repetitive/hallucinated sequences) and reliance on genomic context; lab validation remains essential; applicability to human ge

QC result: Pass.