Molecular Dynamics Affinity Release Notes

v6.5.8 April 2026

General Notice

  • A Rapid FE-NES (Free Energy Nonequilibrium Switching) toggle switch has been added to the Equilibration and Nonequilibrium Switching Floe. When set to On (default: Off), jobs will run shorter MD simulations (1.5 ns) and fewer alchemical switches (20) to boost speed by 2x and throughput by 3x. Typical accuracy degradation is small (~0.2 kcal/mol).

  • Direct experimental affinity ingestion is now available (e.g., as originally provided in input .sdf files).

  • Experimental affinities may now be provided in concentration and log units (e.g., as \(IC_{50}\) and \(pIC_{50}\)).

  • The efficiency of MD simulations has been enhanced by optimizing cycling parameters in production simulations. Most MD-based Floes should now complete more quickly and incur reduced cloud compute costs (>10% improvement for standard MD).

  • Memory allocations for several cubes have been increased to better accommodate large-scale MD and NES simulations.

  • Primary input parameters have been moved to top sections of job forms in many Floes.

  • The package has been updated to run on Python 3.12 in order to align the software with modern runtime environments.

  • The package now pins oemdtoolbox 2.13.5, which includes bug fixes and compatibility with OpenMM ≥ 8. It also pins orionmdcore 2.6.1, which includes core dependency updates.

Floe Updates

  • Several minor updates have been made in the Edge Mapper for RBFE calculations Floe: Empty string hub inputs are now properly considered as unspecified; the Floe checks whether the selected map_type supports hubs and fails if a hub is specified for non-hub types; Edge scores are now shown as ‘N/A’ instead of 1.0 in the Floe reports when edge scoring is intentionally skipped; mitigated rare conditions in which OELOMAP graphs had many more than 1.5 edges per ligand.

  • Several minor updates have been made in the Nonequilibrium Switching Floe: The default maximum edge limit has increased from 100 to 500, and NES jobs now fail with a clear Floe Failure Report when the edge count exceeds this limit. To help prevent accidental all-to-all edge maps, NES Floes also abort when the input edge map is a complete graph with more than 100 edges. To intentionally run on a fully connected graph above this threshold, set the new force_complete_graph parameter (default: False) to True.

  • All Floe outputs now have default names.

  • Output dataset names and descriptions have been standardized for greater consistency. In particular, the Nonequilibrium Switching Floe now defaults to the output dataset name binding_affinity_output_dataset (was affinity_output_dataset).

  • Direct ingestion of experimental affinity from datasets selected for jobs is enabled in Equilibration and Nonequilibrium Switching, Nonequilibrium Switching, NES Result Utilities, Nonequilibrium Switching Recovery and Compare Experimental Affinity with End Point Analysis Results Floes. In addition, experimental affinities (and associated uncertainties) can also be provided in concentration and log units, e.g. in \(IC_{50}\) and \(pIC_{50}\), respectively.

  • Primary input parameters (required and optional) are placed in the top of the job form of Equilibration and Nonequilibrium Switching, Nonequilibrium Switching, NES Result Utilities, Nonequilibrium Switching Recovery and Compare Experimental Affinity with End Point Analysis Results Floes.

  • The default trajectory saving interval for MD Floes has been increased to 0.025 ns. For denser sampling, set the interval to 0.004 ns. For extended simulations, a larger interval is recommended to avoid excessive trajectory sizes and increased AWS storage cost.

  • The ‘MD API Dataset Converter’ Floe has been removed. This Floe is no longer needed because all MD Floes can automatically convert old datasets.