regLM: Designing Realistic Regulatory DNA with Autoregressive Language Models

Abstract

Designing cis-regulatory DNA elements (CREs) with desired properties is a challenging task with many therapeutic applications. Here, we used autoregressive language models trained on yeast and human putative CREs, in conjunction with supervised sequence-to-function models, to design regulatory DNA with desired patterns of activity. Our framework, regLM, compares favorably to existing CRE design approaches at generating realistic and diverse regulatory DNA, while also providing insights into the cis-regulatory code.

PDF NeurIPSW OpenReview Semantic Scholar

Cite

Text

Lal et al. "regLM: Designing Realistic Regulatory DNA with Autoregressive Language Models." NeurIPS 2023 Workshops: GenBio, 2023.

Markdown

[Lal et al. "regLM: Designing Realistic Regulatory DNA with Autoregressive Language Models." NeurIPS 2023 Workshops: GenBio, 2023.](https://mlanthology.org/neuripsw/2023/lal2023neuripsw-reglm/)

BibTeX

@inproceedings{lal2023neuripsw-reglm,
  title     = {{regLM: Designing Realistic Regulatory DNA with Autoregressive Language Models}},
  author    = {Lal, Avantika and Biancalani, Tommaso and Eraslan, Gökcen},
  booktitle = {NeurIPS 2023 Workshops: GenBio},
  year      = {2023},
  url       = {https://mlanthology.org/neuripsw/2023/lal2023neuripsw-reglm/}
}