Dataset Information

Decoding and Rewiring Promoter Architecture Using Large Language Models and Diffusion Frameworks

ABSTRACT: High-performance promoters are essential tools for precisely regulating gene expres-sion, yet their rational design within the vast combinatorial sequence space remains a major challenge. Here, we present a hybrid framework that integrates a large lan-guage model (LLM) with a diffusion model to enable data-driven and interpretable promoter design. The fine-tuned LLM predicts promoter strength with high accuracy and, through pseudo-sequence mutations, identifies biologically essential core motifs. A diffusion model is then conditioned on these motifs to reconstruct non-core regions and generate complete promoter sequences. We experimentally validated this approach in E. coli by high-throughput barcoded promoter activity sequencing: over 90% of the generated promoters showed measurable activity, and the best variants achieved ap-proximately ∼20-fold higher expression than the benchmark promoter (BBa_J23119). By explicitly coupling interpretability with generative design, this strategy provides a generalizable path to accelerate synthetic biology efforts and advance large-scale regu-latory sequence engineering.

ORGANISM(S): Escherichia coli DH5[alpha]

PROVIDER: GSE316915 | GEO | 2026/01/23

REPOSITORIES: GEO

ACCESS DATA

Dataset's files

Source:

			Action	DRS
		Other

Items per page:

1 - 1 of 1

Similar Datasets

Project description:Background: In prokaryotes, sigma factors are essential for the targeting of the transcription machinery to promoters. Various sigma factors have been described that recognize and bind to specific DNA sequence motifs in promoter sequences. The canonical sigma factor σ70 is commonly involved in transcription of the cell's general housekeeping genes, which is mediated by the conserved σ70 promoter sequence motifs. This study detects and predicts the general σ70-promoter sequences in Lactobacillus plantarum WCFS1 using genome-wide analysis. The accuracy of the transcriptionally-active part of this promoter prediction was subsequently evaluated by correlating locations of predicted promoters with experimentally detected transcription starts sites (TSSs) using high-resolution tiling array transcriptome datasets. Results: To identify σ70-related promoter sequences, we performed a genome-wide sequence motif scan of the L. plantarum WCFS1 genome focusing on the regions upstream of protein-encoding genes. We obtained several highly conserved motifs including those resembling the conserved σ70-promoter consensus. Position weight matrices-based models of the recovered σ70-promoter sequence motif were employed to identify 4746 motifs with significant similarity (p-value < 10-4) to the model-motif in the L. plantarum genome. Genome-wide transcription start site information deduced from whole genome tiling-array transcriptome datasets revealed 936 TSSs that were employed to validate the transcriptionally active fraction of these predicted promoters. In total, 578 predicted promoters were found in proximity (≤ 100 nucleotides) of the identified TSSs, showing a highly significant co-occurrence of predicted promoter and measured TSS (p-value < 10-23). An additional 224 predicted promoters was validated when the significant similarity to the model-motif was applied less strictly (10-4 ≤ p-value < 10-3). Conclusions: High-resolution tiling arrays provide a suitable source for TSS detection at a genome-wide level, and allow experimental verification of in silico-predicted promoter sequence motifs.

Project description:Background: In prokaryotes, sigma factors are essential for the targeting of the transcription machinery to promoters. Various sigma factors have been described that recognize and bind to specific DNA sequence motifs in promoter sequences. The canonical sigma factor M-OM-^C70 is commonly involved in transcription of the cell's general housekeeping genes, which is mediated by the conserved M-OM-^C70 promoter sequence motifs. This study detects and predicts the general M-OM-^C70-promoter sequences in Lactobacillus plantarum WCFS1 using genome-wide analysis. The accuracy of the transcriptionally-active part of this promoter prediction was subsequently evaluated by correlating locations of predicted promoters with experimentally detected transcription starts sites (TSSs) using high-resolution tiling array transcriptome datasets. Results: To identify M-OM-^C70-related promoter sequences, we performed a genome-wide sequence motif scan of the L. plantarum WCFS1 genome focusing on the regions upstream of protein-encoding genes. We obtained several highly conserved motifs including those resembling the conserved M-OM-^C70-promoter consensus. Position weight matrices-based models of the recovered M-OM-^C70-promoter sequence motif were employed to identify 4746 motifs with significant similarity (p-value < 10-4) to the model-motif in the L. plantarum genome. Genome-wide transcription start site information deduced from whole genome tiling-array transcriptome datasets revealed 936 TSSs that were employed to validate the transcriptionally active fraction of these predicted promoters. In total, 578 predicted promoters were found in proximity (M-bM-^IM-$ 100 nucleotides) of the identified TSSs, showing a highly significant co-occurrence of predicted promoter and measured TSS (p-value < 10-23). An additional 224 predicted promoters was validated when the significant similarity to the model-motif was applied less strictly (10-4 M-bM-^IM-$ p-value < 10-3). Conclusions: High-resolution tiling arrays provide a suitable source for TSS detection at a genome-wide level, and allow experimental verification of in silico-predicted promoter sequence motifs. Triplicate hybridizations (technical replicates) of two L. plantarum WCFS1 samples (biological replicates) gathered from a fermentation run at OD600 of 1. Dye swap applied to one of the three technical replicates.

Dataset Information

Decoding and Rewiring Promoter Architecture Using Large Language Models and Diffusion Frameworks

Dataset's files

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets