Pose: a pure-NumPy, chirality-aware, structural engine for building and manipulating the molecular structures of proteins, nucleic acids, and small organic molecules
Pose is a lightweight, pure-NumPy Python library that provides a unified, chirality-aware framework for building and manipulating protein, nucleic acid, and small molecule structures while accurately implementing multiple force fields and scoring functions with performance comparable to established tools like OpenMM, PyRosetta, and AutoDock Vina.
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 This is an AI-generated explanation of a preprint that has not been peer-reviewed.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
Scientists who study the shape and behavior of life's building blocks often face a frustrating reality: to understand a single protein, they must juggle a dozen different software programs. Each program speaks its own language, uses its own file formats, and follows its own set of rules. A researcher might build a molecule in one tool, simulate its movement in another, and then score its stability in a third. Every time they switch, the structure must be translated, and in that translation, subtle errors can creep in. Worse, most of these tools struggle to handle molecules that are mirror images of the natural ones. In nature, proteins are built from left-handed amino acids, but scientists sometimes need to create right-handed versions to make drugs that last longer in the body. Existing software often treats these mirror images as foreign or breaks them entirely, forcing researchers to use workarounds that introduce uncertainty.
Enter Pose, a new tool designed to bring this fragmented process into a single, transparent space. Created by Sari Sabban at King Abdulaziz University, Pose is a software library that can build, move, and analyze proteins, DNA, and small drug-like molecules without needing any complex external engines. It is built entirely on a standard mathematical toolkit called NumPy, meaning it runs on almost any computer without the heavy installation requirements of traditional molecular modeling software. The most significant feature of Pose is its ability to treat left-handed and right-handed molecules with perfect symmetry. In the natural world, a left-handed amino acid and its right-handed mirror image are chemically identical in every way except for their orientation in space. Pose respects this rule by applying a single, consistent set of mathematical tables to both, ensuring that a mirror-image molecule receives the exact same energy score as its original. This consistency allows researchers to mix left-handed and right-handed parts in a single chain without the software getting confused or producing contradictory results.
The researchers tested Pose by comparing its calculations against five established, industry-standard programs. They fed the same molecular structures into Pose and the reference programs to see if the numbers matched. On identical coordinates, Pose reproduced the energy calculations of the reference programs with extreme precision. For the force fields used to simulate physical movement, the difference was often less than one-millionth of a kilojoule per mole, a margin so small it is likely due to the way computers handle decimal points rather than a flaw in the logic. When testing the scoring functions used to predict how well a drug might bind to a protein, Pose matched the results of the reference tools to within a few thousandths of a calorie per mole. Crucially, when the researchers mirrored a structure to create a right-handed version, Pose's energy scores remained unchanged, confirming that the software treats the mirror image exactly as nature does. In contrast, one of the reference programs, PyRosetta, showed significant differences in its scores for mirrored structures, revealing an asymmetry in how it handled the data.
Beyond simply calculating energy, Pose was tested on its ability to rebuild complex structures from scratch. The researchers took the angles between atoms in real proteins and asked Pose to reconstruct the entire shape from those angles alone. The result was a structure that differed from the original by less than one-hundredth of an angstrom, a distance smaller than the width of a single atom. This level of accuracy held true even for a short chain of DNA containing right-handed amino acids, a task that many other tools cannot perform reliably. The software also demonstrated its ability to perform dynamic tasks, such as simulating the movement of atoms over time or finding the most stable arrangement of side chains. While these simulations took longer to run than in the specialized, high-speed reference programs, they produced results that were statistically indistinguishable in the middle of the range. For instance, when docking drug molecules into protein pockets, Pose successfully found the correct position in six out of ten test cases, matching the performance of established tools that are widely used in drug discovery.
The work does not claim to be the fastest way to simulate a molecule; in fact, the authors acknowledge that their pure-software approach is significantly slower than the highly optimized, compiled engines used by large pharmaceutical companies. Instead, the value of Pose lies in its clarity and flexibility. Because every calculation is written as a readable line of code rather than hidden inside a black-box program, a researcher can see exactly how an energy score is derived. They can tweak a single term, swap a rule, or add a new type of molecule without needing to recompile a massive software suite. This transparency makes it an ideal tool for teaching students how molecular forces work, for developing new energy functions, or for designing novel molecules that mix left-handed and right-handed parts. By unifying the handling of different force fields and ensuring that mirror images are treated with mathematical honesty, Pose offers a clean, reliable foundation for the next generation of molecular design.
Drowning in papers in your field?
Get daily digests of the most novel papers matching your research keywords — with technical summaries, in your language.