Design and Testing of Root-Specific Synthetic Promoters by Machine Learning

Synthetic promoters are crucial for precise gene expression in transgenic plants, but their rational design is hindered by the difficulty in identifying functional cis-regulatory elements (CREs). In this study, we aimed to develop a systematic approach for discovering tissue-specific cis-regulatory modules (CRMs) and generating functional synthetic promoters in poplar. We performed extensive transcriptomic analysis across various poplar tissues to obtain categorical labels and detected motifs in all gene promoters using known transcription factor binding site (TFBS) position weight matrices. Informative, tissue-specific TFBSs were predicted using a random forest model. Applying this to a root-specific gene, PopRTS1, we identified putative root-specific CRMs. These CRMs were then used to construct synthetic promoters, which were experimentally validated through rapid infiltration and GUS staining assays across different tissues. We successfully identified a root-specific synthetic promoter, PRTS1. Our findings demonstrate that machine learning can effectively decipher regulatory codes from omics data to predict functional CRMs. This work provides a feasible and systematic method for screening and designing tissue-specific synthetic promoters, offering significant potential for advancing targeted gene expression systems in plant biotechnology.