455: Agentic genomics: the bottleneck moves from code to judgment

Corpas M et al., Cell Genomics - A Perspective arguing that autonomous AI agents which discover, configure and chain bioinformatics operations from natural-language instructions have shifted the bottleneck in computational biology from building pipelines to validating their output. The authors define four necessary conditions for a system to count as agentic, propose a perturbation test that separates genuine runtime decision-making from pre-specified branching, survey four existing systems, and set out a three-tier validation framework in which none of the surveyed systems yet reaches clinical grade. Key terms: agentic genomics, AI agents, validation, skill libraries, equity in genomics.

Study Highlights:
The authors define agentic genomics through four necessary conditions — autonomy, domain constraint via validated skill libraries, iterative refinement, and natural-language mediation — and offer an empirically testable criterion: a system is not agentic if perturbed intermediate outputs fail to change its execution strategy. They distinguish the paradigm from workflow managers like Nextflow and Galaxy, from AutoML, from LLM-assisted scripting, and from general-purpose biomedical copilots. Four systems are surveyed and placed against a proposed three-tier validation framework of research grade, benchmarked and clinical grade; all four currently sit at research grade, and none has published the external multi-site evidence the clinical tier requires. The paper documents a concrete silent failure: in the early weeks of ClawBio, one skill given an empty input file containing no genomic data returned all-normal results including recommended dosages for 51 drugs, a path since closed by community audit. It also reports that a randomized trial of GPT-4-augmented diagnostic reasoning found no significant improvement over conventional resources, 76 percent versus 74 percent with p equal to 0.60, even though the model alone scored well above physicians.

Conclusion:
The central claim is that agentic genomics expands the capacity to produce analyses without expanding the capacity to judge them, so validation frameworks must be formalized and tiered, equity-aware defaults must be engineered rather than aspired to, and regulatory frameworks must accommodate non-deterministic software. The authors argue the paradigm is already in use and advancing, and that the open question is not whether it will be adopted but whether the field will build the standards to make it trustworthy first.

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

Article title:
Agentic genomics: From pipeline automation to autonomous validation

First author:
Corpas M

Journal:
Cell Genomics

DOI:
10.1016/j.xgen.2026.101305

Reference:
Corpas M, Guio H, Fatumo S. Agentic genomics: From pipeline automation to autonomous validation. Cell Genomics. 2026;6(8):101305. doi:10.1016/j.xgen.2026.101305

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 is independent and ad-free — no sponsors, no paywall. If an episode was worth your time, chip in and keep the papers audited and the original songs coming:
❤️ Support monthly: https://buy.stripe.com/cNifZhclVebvagk2JDgEg01
☕ One-time donation: https://donate.stripe.com/7sY4gz71B2sN3RWac5gEg00
▶️ Watch with chapters on YouTube: https://www.youtube.com/@basebybase
More at 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/agentic-genomics-validation-bottleneck

QC:
This episode was checked against the original article PDF and publication metadat...