Designed transcriptional activation domains, over 90 percent functional in yeast
A Berkeley-led team generated 500,000 candidate activation domains, scored them with existing machine learning models including three neural networks, and synthesized 18,750 of the top ones, of which more than 90 percent worked as activation domains in yeast, with some retaining function in fungi, plants and human cells.
- Model
- Existing activation-domain function predictors including three neural networks, plus MADRIGAL, a consensus predictor built from the results
- Field
- Biology
- Date
- 2026-10-30
- Human collaborators
- Marissa A. Zintel, Max V. Staller
Sources
Original work
What was found
Protein design has worked best on structured proteins, where a fold constrains the sequence. Activation domains are the opposite case: they are intrinsically disordered regions of transcription factors that recruit coactivators, with weak structural constraints and poorly defined sequence-to-function rules, which makes them a test of whether a prediction-and-screen loop can design disordered sequence at all. Candidates resembling known activation domains were generated computationally, scored by several extant predictors, and 18,750 high scorers were measured in yeast, of which 16,770 gave clean measurements and over 90 percent functioned. Sequences that several models agreed on were stronger, which is what motivated MADRIGAL, the fast consensus predictor the paper then builds. Designed domains that keep working across fungi, plants and human cells matter for synthetic biology, agriculture and biotechnology, where an activation domain is a reusable part.
Novelty check
Nothing in this registry covered activation domains or designed intrinsically disordered protein: a search of data/entries.json for activation domain, intrinsically disordered, transcription factor and MADRIGAL returned no match, and the protein-design entries are all structured designs, binders, enzymes or antibodies. Europe PMC, 2026-11-09, the bioRxiv query behind this lead returned 3 preprints in the window, of which this is the only designed disordered-sequence result. The paper's own novelty framing is that structured protein design has advanced while disordered design has not, with activation domains as one of the oldest known classes of functional disordered sequence.
Caveats and known objections
A bioRxiv preprint, not peer reviewed, and the validation is the authors' own yeast assay with no independent replication, hence author verified. The cross-species results are described as a number of designs retaining function rather than a measured rate, so the headline over 90 percent figure is the yeast assay only. The high hit rate is a property of sequences selected by predictors that were already trained on known activation domains, so it measures the design loop, not an unconstrained generative model. The preprint names no AI disclosure statement; the machine learning role is read off the abstract's own description of generating candidates and predicting function with extant models, and the paper reports a patent application filed alongside the work. Autonomy is search-scaffold rather than ai-led: humans framed the generate-and-screen loop and the models scored candidates inside it.
No independent check is recorded on this entry yet.
Reading the primary source closely enough to say whether it supports the claim counts as a check, and you are credited on the entry.
Or on GitHub: submit a check challenge the grade send a correction or send a pull request
Flag this for triage
Signals order the review queue and nothing else. They are never published, and they never move a grade: that takes a citation.
Entry history (1 event)
- AddedEntered the registry graded Author verified and Search scaffold.
Entries are never deleted. A grade that does not hold up is downgraded on the record, with the reason beside it.
Graded author verified for verification and search scaffold for autonomy. What these mean.
Cite this entry
whataifound.org. (2026). Designed transcriptional activation domains, over 90 percent functional in yeast. whataifound.org: A Registry of AI Scientific and Mathematical Discoveries. https://whataifound.org/finding/2026-10-30-designed-activation-domains
BibTeX
@misc{whataifound-universityofcaliforniaberkeley-2026-domains,
title = {Designed transcriptional activation domains, over 90 percent functional in yeast},
author = {{whataifound.org}},
year = {2026},
howpublished = {whataifound.org: A Registry of AI Scientific and Mathematical Discoveries},
note = {Result by University of California, Berkeley. Verification: Author verified. Autonomy: Search scaffold.},
url = {https://whataifound.org/finding/2026-10-30-designed-activation-domains}
}