Controlling gene expression with deep generative design of regulatory DNA

Jan Zrimec; Xiaozhi Fu; Azam Sheikh Muhammad; Christos Skrekas; Vykintas Jauniskis; Nora K. Speicher; Christoph S. Börlin; Vilhelm Verendel; Morteza Haghir Chehreghani; Devdatt Dubhashi; Verena Siewers; Florian David; Jens Nielsen; Aleksej Zelezniak

doi:10.1038/s41467-022-32818-8

Controlling gene expression with deep generative design of regulatory DNA

Jan Zrimec^*, Xiaozhi Fu, Azam Sheikh Muhammad, Christos Skrekas, Vykintas Jauniskis, Nora K. Speicher, Christoph S. Börlin, Vilhelm Verendel, Morteza Haghir Chehreghani, Devdatt Dubhashi, Verena Siewers, Florian David, Jens Nielsen, Aleksej Zelezniak

^*Corresponding author for this work

Randall Centre of Cell & Molecular Biophysics

Research output: Contribution to journal › Article › peer-review

39 Citations (Scopus)

Abstract

Design of de novo synthetic regulatory DNA is a promising avenue to control gene expression in biotechnology and medicine. Using mutagenesis typically requires screening sizable random DNA libraries, which limits the designs to span merely a short section of the promoter and restricts their control of gene expression. Here, we prototype a deep learning strategy based on generative adversarial networks (GAN) by learning directly from genomic and transcriptomic data. Our ExpressionGAN can traverse the entire regulatory sequence-expression landscape in a gene-specific manner, generating regulatory DNA with prespecified target mRNA levels spanning the whole gene regulatory structure including coding and adjacent non-coding regions. Despite high sequence divergence from natural DNA, in vivo measurements show that 57% of the highly-expressed synthetic sequences surpass the expression levels of highly-expressed natural controls. This demonstrates the applicability and relevance of deep generative design to expand our knowledge and control of gene expression regulation in any desired organism, condition or tissue.

Original language	English
Article number	5099
Journal	Nature Communications
Volume	13
Issue number	1
DOIs	https://doi.org/10.1038/s41467-022-32818-8
Publication status	Published - Dec 2022

Access to Document

10.1038/s41467-022-32818-8

Cite this

@article{7adf1f99f59c44b3a13ab940fadcb1b4,

title = "Controlling gene expression with deep generative design of regulatory DNA",

abstract = "Design of de novo synthetic regulatory DNA is a promising avenue to control gene expression in biotechnology and medicine. Using mutagenesis typically requires screening sizable random DNA libraries, which limits the designs to span merely a short section of the promoter and restricts their control of gene expression. Here, we prototype a deep learning strategy based on generative adversarial networks (GAN) by learning directly from genomic and transcriptomic data. Our ExpressionGAN can traverse the entire regulatory sequence-expression landscape in a gene-specific manner, generating regulatory DNA with prespecified target mRNA levels spanning the whole gene regulatory structure including coding and adjacent non-coding regions. Despite high sequence divergence from natural DNA, in vivo measurements show that 57% of the highly-expressed synthetic sequences surpass the expression levels of highly-expressed natural controls. This demonstrates the applicability and relevance of deep generative design to expand our knowledge and control of gene expression regulation in any desired organism, condition or tissue.",

author = "Jan Zrimec and Xiaozhi Fu and Muhammad, {Azam Sheikh} and Christos Skrekas and Vykintas Jauniskis and Speicher, {Nora K.} and B{\"o}rlin, {Christoph S.} and Vilhelm Verendel and Chehreghani, {Morteza Haghir} and Devdatt Dubhashi and Verena Siewers and Florian David and Jens Nielsen and Aleksej Zelezniak",

note = "Funding Information: We thank Filip Buric and Sandra Viknander for technical discussions and critical comments as well as Benjamin Heineike, Kate Cambell, and Simran Aulakh for proofreading and providing critical feedback on the manuscript. We gratefully acknowledge the NVIDIA Corporation for supporting this research as well as the Chalmers Center for Computational Science and Engineering (C3SE) and the Swedish National Infrastructure for Computing (SNIC) for providing computational resources. Mikael {\"O}hman and Thomas Svedberg at C3SE are acknowledged for technical assistance. The study was supported by SciLifeLab funding (A.Z.), Swedish Research council (Vetenskapsr{\aa}det) starting grant no. 2019-05356 (A.Z.), BigData@Chalmers funding initiative (Area of Advance ICT) (A.Z.), Marius Jakulis Jason foundation (A.Z.), Slovenian Research Agency (ARRS) grant no. J2-3060 (J.Z.), Public Scholarship, Development, Disability, and Maintenance Fund of the Republic of Slovenia grant no. 11013-9/2021-2 (JZ) and EU Horizon 2020 research and innovation program under the Marie Sk{\l}odowska-Curie grant agreement no. 722 287 (C.S.B.). Computing resources at the Chalmers Center for Computational Science and Engineering (C3SE) were partially funded by the Swedish Research Council through grant agreement no. 2018-05973 (A.Z.). Publisher Copyright: {\textcopyright} 2022, The Author(s).",

year = "2022",

month = dec,

doi = "10.1038/s41467-022-32818-8",

language = "English",

volume = "13",

journal = "Nature Communications",

issn = "2041-1723",

publisher = "Nature Publishing Group",

number = "1",

}

TY - JOUR

T1 - Controlling gene expression with deep generative design of regulatory DNA

AU - Zrimec, Jan

AU - Fu, Xiaozhi

AU - Muhammad, Azam Sheikh

AU - Skrekas, Christos

AU - Jauniskis, Vykintas

AU - Speicher, Nora K.

AU - Börlin, Christoph S.

AU - Verendel, Vilhelm

AU - Chehreghani, Morteza Haghir

AU - Dubhashi, Devdatt

AU - Siewers, Verena

AU - David, Florian

AU - Nielsen, Jens

AU - Zelezniak, Aleksej

N1 - Funding Information: We thank Filip Buric and Sandra Viknander for technical discussions and critical comments as well as Benjamin Heineike, Kate Cambell, and Simran Aulakh for proofreading and providing critical feedback on the manuscript. We gratefully acknowledge the NVIDIA Corporation for supporting this research as well as the Chalmers Center for Computational Science and Engineering (C3SE) and the Swedish National Infrastructure for Computing (SNIC) for providing computational resources. Mikael Öhman and Thomas Svedberg at C3SE are acknowledged for technical assistance. The study was supported by SciLifeLab funding (A.Z.), Swedish Research council (Vetenskapsrådet) starting grant no. 2019-05356 (A.Z.), BigData@Chalmers funding initiative (Area of Advance ICT) (A.Z.), Marius Jakulis Jason foundation (A.Z.), Slovenian Research Agency (ARRS) grant no. J2-3060 (J.Z.), Public Scholarship, Development, Disability, and Maintenance Fund of the Republic of Slovenia grant no. 11013-9/2021-2 (JZ) and EU Horizon 2020 research and innovation program under the Marie Skłodowska-Curie grant agreement no. 722 287 (C.S.B.). Computing resources at the Chalmers Center for Computational Science and Engineering (C3SE) were partially funded by the Swedish Research Council through grant agreement no. 2018-05973 (A.Z.). Publisher Copyright: © 2022, The Author(s).

PY - 2022/12

Y1 - 2022/12

N2 - Design of de novo synthetic regulatory DNA is a promising avenue to control gene expression in biotechnology and medicine. Using mutagenesis typically requires screening sizable random DNA libraries, which limits the designs to span merely a short section of the promoter and restricts their control of gene expression. Here, we prototype a deep learning strategy based on generative adversarial networks (GAN) by learning directly from genomic and transcriptomic data. Our ExpressionGAN can traverse the entire regulatory sequence-expression landscape in a gene-specific manner, generating regulatory DNA with prespecified target mRNA levels spanning the whole gene regulatory structure including coding and adjacent non-coding regions. Despite high sequence divergence from natural DNA, in vivo measurements show that 57% of the highly-expressed synthetic sequences surpass the expression levels of highly-expressed natural controls. This demonstrates the applicability and relevance of deep generative design to expand our knowledge and control of gene expression regulation in any desired organism, condition or tissue.

AB - Design of de novo synthetic regulatory DNA is a promising avenue to control gene expression in biotechnology and medicine. Using mutagenesis typically requires screening sizable random DNA libraries, which limits the designs to span merely a short section of the promoter and restricts their control of gene expression. Here, we prototype a deep learning strategy based on generative adversarial networks (GAN) by learning directly from genomic and transcriptomic data. Our ExpressionGAN can traverse the entire regulatory sequence-expression landscape in a gene-specific manner, generating regulatory DNA with prespecified target mRNA levels spanning the whole gene regulatory structure including coding and adjacent non-coding regions. Despite high sequence divergence from natural DNA, in vivo measurements show that 57% of the highly-expressed synthetic sequences surpass the expression levels of highly-expressed natural controls. This demonstrates the applicability and relevance of deep generative design to expand our knowledge and control of gene expression regulation in any desired organism, condition or tissue.

UR - http://www.scopus.com/inward/record.url?scp=85136965773&partnerID=8YFLogxK

U2 - 10.1038/s41467-022-32818-8

DO - 10.1038/s41467-022-32818-8

M3 - Article

C2 - 36042233

AN - SCOPUS:85136965773

SN - 2041-1723

VL - 13

JO - Nature Communications

JF - Nature Communications

IS - 1

M1 - 5099

ER -

Controlling gene expression with deep generative design of regulatory DNA

Abstract

Access to Document

Other files and links

Fingerprint

Cite this