⚛️ biophysics

BENDER: A Cross-taxon IDP Simulation Database Reveals Conserved Sequence-Ensemble Laws Across the Tree of Life

The paper introduces BENDER, a cross-taxon simulation database of over 11,000 intrinsically disordered protein sequences that demonstrates the reliability of current force fields across the tree of life and reveals conserved physical laws governing sequence-ensemble relationships, such as hub topology in contact networks, beyond human-specific evolutionary peculiarities.

Original authors: Velasquez, J., Rahman, T.

Published 2026-08-19

📖 3 min read☕ Coffee break read

CC BY 4.0

Original authors: Velasquez, J., Rahman, T.

Original paper licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). ⚕️ This is an AI-generated explanation of a preprint that has not been peer-reviewed. It is not medical advice. Do not make health decisions based on this content. Read full disclaimer

Life is built from proteins, the molecular machines that carry out nearly every task inside a cell. For decades, scientists have focused on proteins that fold into tight, specific shapes, much like a key fitting into a lock. However, a vast portion of the protein world does not fold at all. These are called intrinsically disordered proteins, and they remain flexible and floppy, drifting through the cell in a constantly shifting cloud of shapes rather than settling into one rigid form. Because they lack a fixed structure, understanding how they behave has been difficult, and most of our knowledge comes from studying human proteins. This leaves a huge gap in our understanding: we do not know if the rules that govern these floppy human proteins apply to the rest of the tree of life, or if they are just a unique quirk of our own species.

A team of researchers has now taken a major step toward filling this gap by creating a massive new resource called BENDER. This is a database containing the sequences of 11,533 disordered proteins drawn from 13 different groups of organisms, ranging from bacteria and plants to animals and viruses. To see how these proteins actually move and interact, the researchers used a powerful computer simulation method to watch them unfold and reform over time. They did not just look at the shapes; they mapped out how the different parts of the proteins touched one another, creating a network of connections that reveals how the protein behaves as a whole. They also checked how well their computer models matched up with real-world structural data to ensure their simulations were accurate.

The results show that the physical laws governing these floppy proteins are remarkably consistent across the entire tree of life. The computer models worked just as well for organisms far removed from humans as they did for human proteins, suggesting that the rules of disorder are universal rather than specific to one branch of evolution. When the researchers compared their findings with a different set of computer rules, they found no hidden biases that would skew the results. Furthermore, they discovered that training their prediction tools on this diverse mix of life forms made them better at guessing the behavior of proteins they had never seen before, including those from viruses.

Perhaps the most striking finding concerns the internal structure of these protein clouds. The researchers found that in every group of organisms they studied, the most connected parts of the protein networks tended to link up with other highly connected parts. This specific pattern, where hubs connect to hubs, appears to be a conserved feature of how disordered proteins are built. It suggests that this way of organizing connections is a fundamental physical property of the molecules themselves, shaped by the laws of physics rather than by the random accidents of evolution. By showing that these patterns hold true from viruses to humans, the study confirms that the physics of disorder is a shared language spoken by all living things.

Technical Summary: BENDER Database and Cross-Taxon IDP Analysis

Problem Statement
While intrinsically disordered proteins (IDPs) and regions are ubiquitous across all kingdoms of life, the computational characterization of their conformational ensembles has been almost exclusively limited to the human proteome. This limitation raises two fundamental questions for the field: first, whether physics-based force fields developed on eukaryotic sequences remain reliable when applied to taxonomically distant organisms; and second, whether the sequence-ensemble relationships derived from these models reflect conserved physical laws or are merely artifacts of a single evolutionary window (the human proteome).

Methodology
To address these gaps, the authors introduce BENDER, a comprehensive simulation database comprising 11,533 IDP sequences spanning 13 distinct taxonomic groups. The methodology involves:

Simulation: All sequences were simulated using the CALVADOS 2 molecular dynamics force field.
Annotation: The dataset includes ensemble-level geometric properties and novel contact-network properties.
Specific Metrics: Per-sequence $\pi$ - $\pi$ and cation- $\pi$ contact frequencies were calculated to link structural features with phase-separation propensity.
Validation Strategy: The study employs a rigorous validation framework involving:
- Comparison against an orthogonal structural reference across the full dataset.
- Evaluation of "held-out" taxa to test generalizability.
- Direct comparison with a second, independently parameterized force field to check for systematic scaling-exponent biases.
- Testing of cross-taxon training data on two independent prediction architectures to assess out-of-distribution (OOD) performance.
- Analysis of ensemble contact-network global efficiency on held-out viral sequences.

Key Contributions

BENDER Dataset: The creation of the first large-scale, cross-taxon IDP simulation database, moving beyond the human-centric focus of previous resources.
Force Field Validation: Empirical evidence demonstrating that CALVADOS 2 ensembles agree strongly with orthogonal structural references across diverse taxa, including those not seen during training.
Network Topology Discovery: The identification of positive degree assortativity as a conserved feature across all analyzed taxonomic groups.

Results

Force Field Reliability: CALVADOS 2 ensembles showed strong agreement with structural references across the entire dataset. Crucially, held-out taxa performed above the dataset median, indicating robustness beyond the training distribution.
Bias Absence: Comparisons with a second, independently parameterized force field revealed no systematic scaling-exponent bias, suggesting the observed trends are not artifacts of a specific force field parameterization.
Predictive Power: Training on cross-taxon data improved out-of-distribution ensemble prediction in two independent architectures. Furthermore, ensemble contact-network global efficiency was found to be accurately predictable from sequence alone, even for held-out viral sequences.
Conserved Topology: Positive degree assortativity was observed consistently across all taxonomic groups. This suggests that hub topology in disordered protein contact networks is a conserved physical feature encoded by the sequence, rather than an evolutionary contingency specific to certain lineages.

Significance
The paper claims that these findings establish that the physical laws governing IDP conformational ensembles are conserved across the tree of life. By demonstrating that sequence-ensemble relationships are not peculiar to a single evolutionary window (humans) but are instead generalizable physical principles, the study validates the use of current physics-based force fields for non-human organisms. The identification of conserved network topologies further implies that the structural organization of disorder is a fundamental, sequence-encoded property of proteins, transcending specific evolutionary histories.

Technical Summary: BENDER Database and Cross-Taxon IDP Analysis

More like this