Understanding the genetic regulatory code that governs gene expression is a primary, yet challenging aspiration in molecular biology that opens up possibilities to cure human diseases and solve biotechnology problems. However, the fundamental question of how each of the individual coding and non-coding regions of the gene regulatory structure interact and contribute to the mRNA expression levels remains unanswered. Considering that all the information for gene expression regulation is already present in living cells, here we applied deep learning on over 20,000 mRNA datasets to learn the genetic regulatory code controlling mRNA expression in 7 model organisms ranging from bacteria to human. We show that in all organisms, mRNA abundance can be predicted directly from the DNA sequence with high accuracy, demonstrating that up to 82% of the variation of gene expression levels is encoded in the gene regulatory structure. Coding and non-coding regions carry both overlapping and orthogonal information and additively contribute to gene expression levels. By searching for DNA regulatory motifs present across the whole gene regulatory structure, we discover that motif interactions can regulate gene expression levels in a range of over three orders of magnitude. The uncovered co-evolution of coding and non-coding regions challenges the current paradigm that single motifs or regions are solely responsible for gene expression levels. Instead, we propose a holistic system that spans all regions of the gene structure and is required to analyse, understand, and design any future gene expression systems.