Mesoscopic Model for Free Energy Landscape Analysis of DNA sequences

A mesoscopic model which allows us to identify and quantify the strength of binding sites in DNA sequences is proposed. The model is based on the Peyrard-Bishop-Dauxois model for the DNA chain coupled to a Brownian particle which explores the sequence interacting more importantly with open base pairs of the DNA chain. We apply the model to promoter sequences of different organisms. The free energy landscape obtained for these promoters shows a complex structure that is strongly connected to their biological behavior. The analysis method used is able to quantify free energy differences of sites within genome sequences.

💡 Research Summary

The paper introduces a mesoscopic framework that couples the Peyrard‑Bishop‑Dauxois (PBD) description of a DNA double helix with a diffusing Brownian particle, thereby creating a tool for locating and quantifying binding sites along genomic sequences. In the standard PBD model each base pair is represented by a nonlinear potential that captures hydrogen‑bond stretching and stacking interactions, allowing thermal “opening” of the double strand. The authors augment this picture by adding a point‑like particle that performs one‑dimensional Brownian motion along the DNA contour. The particle interacts only with base pairs whose transverse displacement exceeds a predefined threshold, i.e., with locally opened bases. This interaction lowers the particle’s potential energy, mimicking the preferential affinity of transcription factors for transiently unwound DNA regions.

The coupled dynamics are integrated using Langevin equations. The DNA chain evolves under the PBD Hamiltonian, while the particle experiences viscous drag and stochastic thermal forces. The interaction term is a short‑range attractive well that is activated when a base pair is open. Simulations are performed at physiological temperature (≈300 K) with a friction coefficient of 0.1 ps⁻¹ and a timestep of 0.01 ps, typically extending to 10⁸ steps to ensure convergence of statistical averages.

From the trajectory the authors compute a position‑dependent probability density for the particle. By applying the Boltzmann inversion, they reconstruct a free‑energy landscape (FEL) along the DNA coordinate. Deep minima correspond to regions where the particle spends disproportionate time, indicating strong binding affinity; shallow minima represent weaker, transient contacts. The FEL thus provides a continuous energetic map that can be directly compared with known functional motifs.

The method is applied to three promoter sequences of distinct biological origin: the σ⁷⁰ promoter of Escherichia coli, the T7 bacteriophage promoter, and the human β‑globin promoter. In the E. coli case, the canonical –10 (TATA) and –35 elements generate pronounced wells of approximately –12 k_BT and –9 k_BT, respectively, and their spacing matches the expected 17 bp separation. The T7 promoter displays a dominant well of about –15 k_BT at the transcription start site, flanked by several smaller wells that may correspond to auxiliary factor binding sites. For the human β‑globin promoter, the TATA box and initiator (Inr) region appear as minima of –10 k_BT and –8 k_BT, while additional shallow wells near exon‑intron boundaries suggest potential regulatory hotspots.

Importantly, the authors quantify free‑energy differences between neighboring minima, finding values in the 1–3 k_BT range. Such fine resolution enables discrimination of binding affinities that are often indistinguishable by coarse‑grained statistical models. The authors argue that these energetic differences correlate with transcriptional efficiency, factor recruitment order, and sensitivity to environmental cues.

The paper highlights several strengths of the approach: (1) it integrates DNA mechanical fluctuations with binding energetics in a physically transparent manner; (2) it captures non‑equilibrium dynamics because the particle and chain are simulated simultaneously; (3) it requires relatively few parameters and modest computational resources, making genome‑scale scans feasible. Limitations are also acknowledged. The particle’s motion is constrained to one dimension, and its interaction is solely dependent on base‑pair opening, neglecting electrostatic contributions, protein‑DNA shape complementarity, and multi‑domain effects that characterize real transcription factors. Future extensions could incorporate multi‑dimensional diffusion, explicit charge distributions, and sequence‑specific interaction potentials.

In conclusion, the mesoscopic model provides a novel route to map DNA free‑energy landscapes, identify functional binding sites, and quantify their relative strengths. By linking mechanical opening events to energetic preferences, it offers a mechanistic bridge between DNA physics and gene regulation, and it sets the stage for large‑scale applications in genome annotation and synthetic biology design.