Machine learning guided design of strong transcriptional activation domains
- 1Department of Molecular and Cell Biology, University of California Berkeley, Berkeley, CA 94720; Center for Computational Biology, University of California Berk
Abstract
Artificial intelligence has accelerated the design of structured proteins, but it remains difficult to design functional intrinsically disordered proteins because of weak structural constraints and poorly defined sequence-to-function relationships1. Activation domains (ADs), transcription factor regions that recruit coactivators, are an excellent test case for the de novo design of intrinsically disordered sequences. ADs are one of the oldest groups of functional intrinsically disordered sequences2 and designed ADs have applications in synthetic biology, biotechnology, and agriculture. We computationally generated 500,000 candidate sequences that resembled known ADs and predicted AD function with extant machine learning models, including three neural networks. From high-scoring sequences, we selected 18,750 for experimental validation in yeast. Surprisingly, >90% of the 16,770 well-measured sequences function as ADs. Sequences predicted to be active by multiple models were stronger, motivating development of a fast and accurate consensus AD predictor, MADRIGAL. A number of our designed ADs retain functionality across fungi, plants, and human cells. This work illustrates a generalizable paradigm for incorporating sequence-to-function models into the design loop to efficiently generate functional intrinsically disordered sequences.
Related
- Physics-guided explainable machine learning for predicting temporal scour evolution around hybrid spur dikes under unsteady flood hydrographs
- Machine learning classifies subcellular Golgi morphology from imaging flow cytometry during inflammatory activation: an exploratory study
- Machine Learning-Guided Classification of Druggable Pockets and Phylogenetic Druggability Transfer Across the Human Kinome
- Machine-learning-guided enzyme discovery and redox-system engineering for efficient production of the nylon monomer methyl 12-aminododecanoate in Escherichia coli
- Transcriptomics-guided serum podoplanin and interpretable machine learning for early risk stratification of necrotizing enterocolitis in preterm infants
BibTeX
@misc{zintel2026machine,
title = {{Machine learning guided design of strong transcriptional activation domains}},
author = {M. A. Zintel and C. J. LeBlanc and S. R. Kotha and M. Soriano and R. J. Bansal and D. Aguilera and T. Majidi and I. Ornelas and L. D. Kirkpatrick and A. Udupa and G. Hu and P. Agarwal and A. W. Y. Lam and J. K. Nunez and P. M. Shih and R. Sozzani and L. C. Strader and M. V. Staller},
howpublished = {bioRxiv},
year = {2026},
month = oct,
doi = {10.64898/2026.10.30.763198},
url = {https://www.biorxiv.org/content/10.64898/2026.10.30.763198v1},
note = {CC BY-NC},
}
RIS
TY - UNPB
TI - Machine learning guided design of strong transcriptional activation domains
AU - M. A. Zintel
AU - C. J. LeBlanc
AU - S. R. Kotha
AU - M. Soriano
AU - R. J. Bansal
AU - D. Aguilera
AU - T. Majidi
AU - I. Ornelas
AU - L. D. Kirkpatrick
AU - A. Udupa
AU - G. Hu
AU - P. Agarwal
AU - A. W. Y. Lam
AU - J. K. Nunez
AU - P. M. Shih
AU - R. Sozzani
AU - L. C. Strader
AU - M. V. Staller
PY - 2026
DA - 2026/10/30
PB - bioRxiv
DO - 10.64898/2026.10.30.763198
UR - https://www.biorxiv.org/content/10.64898/2026.10.30.763198v1
AB - Artificial intelligence has accelerated the design of structured proteins, but it remains difficult to design functional intrinsically disordered proteins because of weak structural constraints and poorly defined sequence-to-function relationships1. Activation domains (ADs), transcription factor regions that recruit coactivators, are an excellent test case for the de novo design of intrinsically disordered sequences. ADs are one of the oldest groups of functional intrinsically disordered sequences2 and designed ADs have applications in synthetic biology, biotechnology, and agriculture. We computationally generated 500,000 candidate sequences that resembled known ADs and predicted AD function with extant machine learning models, including three neural networks. From high-scoring sequences, we selected 18,750 for experimental validation in yeast. Surprisingly, >90% of the 16,770 well-measured sequences function as ADs. Sequences predicted to be active by multiple models were stronger, motivating development of a fast and accurate consensus AD predictor, MADRIGAL. A number of our designed ADs retain functionality across fungi, plants, and human cells. This work illustrates a generalizable paradigm for incorporating sequence-to-function models into the design loop to efficiently generate functional intrinsically disordered sequences.
ER -