2026 August Monthly Release

2026 Aug Monthly Update

Release Notes

1. Website and Documentation Synchronization

   The MAPLE official website has been synchronized with this release. New and expanded documentation covers the ParmFit dispatcher and its Correction, NCAA, and MetalAA routes, Direct MaxFlux transition-state searches, and FeNNol/FeNNix-Bio1 model configuration. @AmiHaruka

   Installation, calculator, coordinate and input/output guidance, examples, and website navigation have also been updated to keep dependency requirements, supported interfaces, and documented workflows aligned with the current code behavior.

2. ParmFit Force-Field Parameterization Workflows

   MAPLE introduces a new parmfit dispatcher for force-field parameterization workflows. The new implementation covers ligand parameter correction, metal-site template construction, and non-canonical amino-acid template generation through Correction, MetalAA, and NCAA routes. QM and MLIP stages can be combined where configured, with structured reports and reusable work directories for charge fitting, bonded-parameter refinement, and downstream export.

   ParmFit also supports torsion-fitting workflows, QM interfaces for Gaussian and ORCA, charge fitting, Seminario/mSeminario bonded refinement, and Amber/GROMACS parameter-file generation. The included examples cover Toluene correction, 6JQH NCAA template generation, 4ZF6 and 1OKL metal-site parameterization, and TorsionFit artifacts. @AmiHaruka

3. PDB Input, Output, and Template Matching

   MAPLE now provides native PDB reading and writing for the relevant molecular and reaction workflows. PDB inputs can be resolved from MAPLE input files, while PDB metadata and output templates are preserved across supported transition-state and trajectory workflows.

   ParmFit adds structure-aware residue and atom matching for protein, metal-site, and non-canonical-residue contexts. The reader and matching layers provide diagnostics for residue selection, alternate locations, atom naming, and template compatibility before parameterization proceeds. @AmiHaruka

4. Direct MaxFlux Transition-State Search

   A new Direct MaxFlux (DMF) transition-state search method is available through #ts(method=dmf). The workflow constructs an initial path with FB-ENM, solves the double-ended variational problem, identifies the maximum-flux path point, and can refine the resulting transition-state guess with PRFO.

   DMF examples are included for standard MLIP workflows and FeNNix-based calculations, together with path, maximum-point, trajectory, and refined transition-state outputs. @AmiHaruka

5. FeNNol Calculator and FeNNix-Bio1 Foundation Models

   MAPLE adds a FeNNol calculator backend with support for the fennol, fennix-bio1s, and fennix-bio1m model names. The built-in FeNNix-Bio1 model files are resolved from MAPLE's model directory, while external .fnx models can be supplied through an explicit model_path.

   The backend exposes energy, forces, analytic Hessians, and Hessian-vector products through MAPLE's calculator interface. Total-charge input, energy-unit conversion, model inspection, and runtime validation are integrated with the shared calculator finalization path. @AmiHaruka

6. Explicit-Solvent Density Reference Compatibility

   Explicit-solvent handling now preserves the legacy water-box density reference used by the MAPLE solvation workflow, maintaining compatibility with established density-based solvent-cluster generation. @AxiEj

Notes

• ParmFit workflows require route-specific external software and parameter data, including AmberTools/tleap with a configured $AMBERHOME and, where selected, Gaussian or ORCA. QM, MLIP, and Amber parameterization stages must be configured in the user's environment.
• Direct MaxFlux requires an IPOPT-compatible Python interface such as cyipopt.
• FeNNol/FeNNix workflows require the corresponding JAX and Flax runtime. Models containing equivariant MACE modules additionally require e3nn_jax.
• MAPLE exposes FeNNol as a non-periodic calculator in this release and requires singlet multiplicity (mult=1). The built-in FeNNix-Bio1 models accept total molecular charge; charge handling for external .fnx models depends on their architecture.
• The core MAPLE installation does not install these external runtimes automatically. Validate model files, external program versions, parameter data, and generated topology files in the target environment before production calculations.
• Amber and other third-party parameter data remain subject to the licensing and redistribution terms of their respective upstream projects.

Contributors

@AmiHaruka, @AxiEj, @Hsuchein