Summary:

Introduces a complete, multi-stage sidechain placement workflow (workflows::place), providing a powerful and configurable method for conformational sampling. The workflow is orchestrated by a new, underlying engine module, which manages configuration, state, and task execution. This feature necessitated a significant refactoring of the forcefield and rotamer modules to support a more robust and performant architecture. Key changes include making the rotamer library self-contained with topology/charge information, caching VDW parameters directly on atoms, and implementing a canonical sorting system for reproducible I/O.

Changes:

• Implemented the Placement Workflow (workflows::place):

  • A new top-level workflow orchestrates the entire sidechain placement process through distinct, observable phases:
    1. Setup: Loads all necessary resources like the forcefield and rotamer library.
    2. EL Pre-computation: Caches Empty Lattice energies for all possible rotamers against the static backbone.
    3. Initialization: Places initial ground-state rotamers to establish a baseline.
    4. Clash Resolution: Iteratively resolves the most severe steric clashes using doublet optimization.
    5. Simulated Annealing (Optional): Performs a global search to escape local energy minima.
    6. Final Refinement: Conducts a final pass of singlet optimization to refine the best solutions.

• Refactored Core engine Module:

  • This module provides the foundational components to support the workflow:
  • config.rs: Fluent builders for creating type-safe configurations (PlacementConfig, DesignConfig, etc.).
  • state.rs: An OptimizationState manager using a BinaryHeap to efficiently track the top N solutions.
  • context.rs: A Context struct to centralize access to the system, configuration, and libraries, featuring a robust residue selection resolver (including ligand-binding sites via k-d tree).
  • tasks/: A suite of modular, parallelizable tasks for el_energy, doublet_optimization, clash_detection, and total_energy calculations.
  • placement.rs: Logic for placing rotamers onto a backbone using the Kabsch alignment algorithm.

• Refactored forcefield, rotamer, and I/O Systems:

  • Decoupled Topology/Charges: Forcefield is now leaner. Topology and charge information is self-contained within the RotamerLibrary definitions, simplifying the data flow.
  • Cached VDW Parameters: Atom structs now cache their VDW parameters (CachedVdwParam), significantly speeding up energy calculations by avoiding repeated lookups.
  • Enhanced Rotamer Library: The rotamer definition now includes internal bond information. The library loader is responsible for the full parameterization of each rotamer upon loading.
  • Canonical I/O Sorting: A new io::sorting module ensures that output files have a deterministic, canonical atom order, making them clean and easily comparable.

• Added Dependencies and Features:

  • Integrated rayon under a parallel feature flag for high-performance, parallel task execution.
  • Added kiddo for efficient k-d tree spatial lookups.
  • Added itertools, rand, and tracing for advanced iteration, sampling, and logging.