Going 15–30× Faster in Relative Binding Free Energy Calculations

A robust new RBFE approach delivers order-of-magnitude speedups without affecting accuracy, unlocking reliable free-energy predictions for real drug discovery workflows

The development and testing of compounds in drug discovery is difficult and expensive, so medicinal chemists use different approaches to tackle this. In silico methods reduce time and cost by identifying promising molecules before synthesis or in vitro testing. Relative Binding Free Energy (RBFE) is a key element in this first screening.

However, RBFE methods are fragile for due diligence projects, which greatly restricts their use. While simulation runtimes are typically stable, accuracy degrades rapidly as molecule modifications grow larger, leading to poor convergence and unreliable results. In practice, this means expensive simulations that often fail to justify the time and compute invested once you move beyond small, incremental changes.

A new RBFE methodology now addresses these challenges head-on, delivering a 15-30× speedup in calculations while preserving state-of-the-art accuracy. The team christened the new method, Dual-LAO with “Dual” for Dual-topology with a dual set of restraints, while LAO stands for Lambda-ABF-OPES. This fresh approach is more robust and reliable when handling chemical transformations, allowing for widespread adoption.  

The Leap in RBFE Performance, Powered by Speed and Accuracy

The standout gain is speed. Dual-LAO runs roughly 15–30× faster than standard RBFE workflows, shrinking simulations from tens of nanoseconds to just a few. Crucially, the speed-up doesn’t come at the expense of accuracy. Across a wide range of test cases, the method delivers errors of roughly 0.5–0.6 kcal/mol, squarely in line with established RBFE approaches.

Furthermore, the calculations are more robust. RBFE has traditionally been problematic for challenging molecular transformations. For example, with scaffold changes, buried water displacements, net charge changes, and ring opening and closure. 

The new method overcomes all these hurdles and is a major leap towards having a dependable, reliable tool. 

Lambda-ABF-OPES leans on adaptive bias and metadynamics-based sampling to help the system cross free energy barriers , drastically accelerating convergence.

With a dual-topology ligand setup, both the starting and ending ligands stay present throughout the simulation, which avoids the tedious dummy-atom constructions of single topology. Dual-DBC restraints prevent drift by keeping both ligands in the binding site throughout the transformation.

Streamlined Workflow and Higher Throughput

The new method improves drug discovery workflow. The up to 30x acceleration means accurate computations can be done within hours, which allows teams to iterate more quickly on design ideas. 

The researchers showed that one simulation could handle several ligand transformations at the same time. This could allow medicinal chemists to evaluate hosts of interactions in parallel, which further streamlines current workflows. 

The new approach is a step change for the pharmaceutical industry. It further expands the range of applicability of such methods in the drug discovery process. Companies can use RBFE predictions as routine tools in decision-making and prioritization with minimal delay in the overall process. 

Greater Accuracy and Scope for Chemical Changes

Greater accuracy and scope to solve difficult chemical changes allows drug researchers to have higher confidence when making go or no-go decisions on their drug analogs.

As the new method handles previously too challenging transformations to be tackled, chemists can be much bolder in their drug design ideas. For example, they can now jump between distinct scaffolds or adding new charged groups, which lets them navigate a new space of chemistry  and potentially discover novel drug candidates, which would have previously been too risky to consider. 

Finally, the Dual-LAO approach can also be applied to other biomolecular systems, like small-molecule binding to RNA, provided that they have structural information of their target of interest.

Yet, RBFE calculations strongly rely on the protein's ligands initial binding pose no matter how powerful the Dual-LAO approach is. 

Still, we are now emerging from a laborious and intense process into one that slides seamlessly into real-world project timelines by making free energy predictions more reliable and faster than ever. The time has come for researchers to boldly explore new domains.

Credits

Massive Congratulations to the authors of the paper this blog is based on: Narjes Ansari, Félix Aviat, Jérôme Hénin, Jean-Philip Piquemal,†,¶ and Louis Lagardère

Read the paper here: https://arxiv.org/abs/2512.17624