DNA Sequence Perplexity Reveals Evolutionarily Conserved Patterns in cis-Regulatory Regions Across Diverse Species

Deciphering cis-regulatory regions in genomes is essential for understanding various physiological processes and pathological mechanisms. Regulatory signatures, namely promoter motifs, transcription factor binding sites, enhancers, GC content, CpG islands, DNA structural motifs, and other cis-regulatory features, are well-established for their roles in transcriptional regulation. However, these features often exhibit species-specific variations, challenging the identification of conserved regulatory principles across different genomes. In this study, we introduce DNA sequence perplexity as an innovative and efficient information-theoretic metric for characterizing cis-regulatory regions. Derived from information theory and natural language processing, perplexity quantifies the complexity and predictability of sequence, offering a motif-independent framework for DNA analysis. By examining transcription and translation start site regions across 1180 species spanning diverse taxa, we demonstrate that cis-regulatory regions consistently exhibit lower perplexity compared to adjacent flanking regions. This trend persists irrespective of taxonomic classification, establishing perplexity as an evolutionarily conserved pattern of regulatory DNA. Additionally, we observe an inverse correlation between perplexity and promoter strength in yeast datasets, suggesting that higher transcriptional outputs are associated with markedly reduced sequence perplexity. Our findings reveal that perplexity may hold valuable insights into the generalizable aspects of cis-regulatory DNA architecture. Integrating this abstraction-based strategy with motif-based approaches and high-throughput functional datasets could enhance its applicability in predictive applications across comparative and functional genomics.