An enhanced sampling algorithm we developed and released as open source, for exploring the many possible 3D shapes a peptide can fold into — published in Physical Chemistry Chemical Physics (Royal Society of Chemistry) and extended in a follow-up preprint, DNCS 1.0.
Peptides and proteins are not rigid objects — the same chain of amino acids can twist and fold into many different 3D shapes, called conformations. Some of these shapes are more stable (lower energy) than others, and figuring out which shapes a molecule is actually likely to adopt is a core problem in structural biology and drug design.
The straightforward way to explore these shapes computationally is random sampling: try a large number of random conformations and see which ones are stable. But random sampling is wasteful — it revisits similar shapes repeatedly and can easily miss entire regions of possibility, especially for intrinsically disordered peptides (IDPs), which don't settle into one fixed shape at all and are notoriously hard to model.
Instead of random sampling, DNCS uses digital nets — sequences of points constructed mathematically (using bitwise XOR operations on binary numbers) so that they cover a space of possibilities evenly, with far less overlap and far fewer wasted samples than random sampling.
In plain terms: imagine mapping every room in a huge building. Random sampling means wandering the corridors hoping to stumble onto each room eventually. DNCS instead walks a planned route that is guaranteed to cover every wing, using a fraction of the steps.
Concretely, DNCS represents a peptide's shape using its dihedral angles — the twist angles along the backbone that determine how it folds — and uses digital nets to generate well-spread samples of those angles. Each sampled conformation is then evaluated with standard molecular force fields (AMBER99SB and AMBER-FB15) and ranked by energy using Boltzmann weighting, so the method reports which shapes are actually plausible, not just which ones were sampled.
The original method was validated on met-enkephalin, a small, notoriously flexible neuropeptide with no single stable shape, and then tested on 42 additional short peptides (3–8 residues) with known crystal structures.
On met-enkephalin specifically, DNCS found a structure 1.64–1.91 Å from the crystal structure (depending on force field) and 1.74 Å from the lowest-energy NMR ensemble structure — in a fraction of the computational time needed by well-tempered metadynamics, a widely used alternative method.
Introduces digital nets conformational sampling, validated on met-enkephalin and 42 short oligopeptides.
Extends the method with an adaptive sampling strategy so it scales to longer, structure-forming peptides — tested on an 11-residue antimicrobial peptide (PDB 6RRO) — and can now also predict secondary structure (helices, coils, turns), not just low-lying conformations.
DNCS 1.0 is released as open-source software under the GNU GPL 3.0 license, so other researchers can use it directly rather than reimplementing the method from the paper.
Abraham Rebairo J., Sam Paul D. and Stephen Arumainathan, "Digital nets conformational sampling (DNCS) – an enhanced sampling technique to explore the conformational space of intrinsically disordered peptides," Phys. Chem. Chem. Phys., 2024. DOI: 10.1039/D4CP01891E.
Abraham Rebairo J., Sam Paul D., Satheeshkumar S. and Stephen Arumainathan, "DNCS 1.0 – Enhanced Conformational Sampling Using Digital Nets Conformational Sampling," ChemRxiv, March 2025. DOI: 10.26434/chemrxiv-2025-wfbtk.
The 2025 preprint has not yet completed peer review. We'll update this page if that changes.
DNCS is open source and ready to use — or talk to us about a custom sampling or modelling problem.
Get in touch