Long-term evolution of regulatory DNA sequences. Part 1: Simulations on global, biophysically-realistic genotype-phenotype maps

Promoters and enhancers are cis-regulatory elements (CREs), DNA sequences that bind transcription factor (TF) proteins to up- or down-regulate target genes. Decades-long efforts yielded TF-DNA interaction models that predict how strongly an individual TF binds arbitrary DNA sequences and how individual binding events on the CRE combine to affect gene expression. These insights can be synthesized into a global, biophysically-realistic, and quantitative genotype-phenotype (GP) map for gene regulation, a “holy grail” for the application of evolutionary theory. A global map provides a rare opportunity to simulate long-term evolution of regulatory sequences and pose several fundamental questions: How long does it take to evolve CREs de novo? How many non-trivial regulatory functions exist in sequence space? How connected are they? For which regulatory architecture is CRE evolution most rapid and evolvable? In this article, the first of a two-part series, we briefly review the pertinent modeling and simulation efforts for a unique system that enables close, quantitative, and mechanistic links between biophysics, as well as systems, synthetic, and evolutionary biology.

💡 Research Summary

This review article surveys recent progress toward building and exploiting global, biophysically realistic genotype‑phenotype (GP) maps for the long‑term evolution of cis‑regulatory DNA sequences such as promoters and enhancers. The authors begin by framing evolution as a three‑part mapping problem—genotype → phenotype → fitness—and emphasizing that while population‑genetic theory is well developed, the genotype‑to‑phenotype step remains the major bottleneck, especially for non‑coding DNA where the sequence space (4^L) is astronomically large.

Two complementary approaches are contrasted. “Local” maps are derived from massively parallel reporter assays (MPRA) that measure expression for tens of thousands of mutant sequences. These provide high‑resolution, quantitative data but only cover a tiny, mutation‑proximal region of sequence space. “Global” maps aim to assign a phenotype to every possible sequence, enabling simulations that start from random DNA and run indefinitely, thereby addressing de‑novo CRE evolution. Existing global models (e.g., House‑of‑Cards, Mount Fuji, NK) are mathematically tractable but biologically simplistic; the challenge is to embed realistic biophysical constraints.

The core of a realistic regulatory GP map is the physical model of transcription‑factor (TF)–DNA binding. TFs recognize short motifs (ℓ≈6–20 bp) with binding energies that can be captured by consensus sequences, position‑weight matrices, or more sophisticated energy matrices that include dinucleotide interactions. Recent MPRA‑derived deep‑learning models can predict up to ~80 % of expression variance across random or designer libraries, but they remain local. To achieve global coverage, the authors propose two extensions. First, the phenotype must be a regulatory function—gene expression as a function of the cellular environment (TF concentrations, signaling states). Second, a “regulatory grammar” must be defined to describe how multiple binding sites, their spacing, orientation, and cooperativity combine to produce the observed function. This grammar acts as an intermediate abstraction that reduces the effective dimensionality of the mapping problem, turning the convolutional nature of TF binding into a tractable computational model.

Thermodynamic frameworks are highlighted as the standard way to translate TF binding energies and concentrations into promoter occupancy, which then maps monotonically onto expression levels. By integrating environment‑dependent TF concentrations, these models can simulate how changes in external conditions shape evolutionary trajectories. The authors also discuss how deep‑learning models trained on MPRA data can be used to fill gaps where experimental measurements are unavailable, thereby extending the global map.

Using this combined framework, several key evolutionary questions become addressable: (1) the expected time for a functional CRE to arise from a non‑functional sequence, (2) how different regulatory architectures (simple prokaryotic promoters versus complex eukaryotic enhancers) affect the speed and accessibility of functional solutions, and (3) the volume and connectivity of functional regions in the full sequence space, i.e., the “regulatory code”. Simulations suggest that enhancers, with many loosely specific TF sites, explore a larger functional volume and evolve more rapidly than tightly constrained promoters.

The review concludes with a roadmap for future work. It calls for richer TF‑DNA binding datasets across diverse conditions, incorporation of non‑equilibrium dynamics (binding/unbinding kinetics, chromatin remodeling), and coupling of the GP map to explicit population‑genetic models that include realistic mutation rates and population sizes. Such advances would enable truly quantitative predictions of long‑term regulatory evolution, bridging the gap between synthetic biology’s design of gene circuits and evolutionary biology’s study of natural regulatory diversity.