Energetic and Safety Signatures of Type I vs Type II LRRK2 Kinase Inhibitors: A Quantum Biochemistry–Guided Reanalysis of Cryo-EM Structures for Parkinson’s Therapeutic Design

1. Background

1.1 Scientific rationale and importance

Leucine-rich repeat kinase 2 (LRRK2) is a heavily pursued drug target for Parkinson’s disease (PD) and regulates pathways such as ciliogenesis, mitophagy, autophagy, and mitochondrial homeostasis. Mutations that increase LRRK2 kinase activity are major genetic drivers of PD, supporting kinase inhibition as a therapeutic strategy.

Despite intense development, translation has been slowed by an incomplete structural–energetic understanding of how different inhibitor classes control LRRK2 conformational states and why potency/selectivity and liabilities differ. Type I and type II inhibitors are both ATP-competitive but are associated with distinct conformational preferences; type II approaches are actively pursued due to concerns with type I inhibitors, including LRRK2 filamentation.

A key recent advance is the release of near-atomic cryo-EM co-structures of LRRK2 bound to a panel of inhibitors: type I (LRRK2-IN-1, GNE-7915) and type II (ponatinib, rebastinib, GZD-824).

These structures provide experimentally validated templates for a structure → energetics → safety framework, analogous to the quantum-biochemistry-guided approach in the template you provided.

1.2 Why static structures are not enough (gap this project addresses)

The cryo-EM models resolve binding poses and key contacts, but three actionable gaps remain for rapid, translational progress:

Energetic causation: The structures reveal contacts (e.g., recurring polar interactions) but do not quantify per-residue energetic contributions that discriminate type I vs type II binding stability and state selection. For type II inhibitors, a recurring pattern includes polar interactions involving A1950, D2017, and E1920, plus direct contact of A2016, consistent with known sensitivity of the A2016T mutation.
Conformational plasticity under realistic dynamics: Type II inhibitors induce rearrangements in the ATP-site, including activation-segment changes and αC helix movement; they occupy space otherwise taken by an activation-segment helix, while Y2018 and a K1906–E1920 salt-bridge change.
Static snapshots do not tell which of these features dominate stability over time.
Safety/ADME liabilities are not integrated with binding physics: LRRK2 inhibitors span scaffolds with very different chemical and predicted safety profiles. An integrated energetic + AI ADME-Tox “fingerprint” can yield design rules to prioritize safer, brain-relevant scaffolds.

1.3 Overall objective (6-month proof-of-concept)

This preregistered project will re-analyze existing public structural datasets (PDB/EMDB coordinates and maps) and generate a comparative energetic atlas of LRRK2 inhibitor binding, integrating MD + quantum biochemistry (per-residue energetics) with consensus AI ADME-Tox to deliver actionable hypotheses and design rules for PD-relevant LRRK2 inhibition.

1.4 Key research questions

What are the dynamic, energetically validated binding modes of type I vs type II inhibitors of LRRK2 using the new cryo-EM templates?
Which residues/secondary-structure elements form energetic “hotspots” that explain class-specific state preference (active-like vs inactive)?
Do residues implicated by cryo-EM contacts (e.g., A2016) emerge as top energetic determinants, and can we rationalize known sensitivity patterns to type II inhibitors?
How do energetic signatures relate to predicted ADME-Tox liabilities, enabling a “physics-informed safety prioritization” of scaffolds?

1.5 Data sources (existing/public datasets)

We will use the published cryo-EM co-structures and associated deposited data, focusing on the inhibitor panel in the paper: type I: LRRK2-IN-1, GNE-7915; type II: ponatinib, rebastinib, GZD-824.

2. Hypothesis

H1 (Energetic class separation): Type I vs type II inhibitors will exhibit distinct per-residue energetic signatures (ΔE_residue patterns) that track their stabilization of active-like vs inactive kinase conformations. This is expected because type I-bound KIN domains are nearly identical to ATP-bound active state (RMSD < 0.5 Å), while type II binding reorganizes the ATP site and αC/activation segment features.
H2 (Hotspot validation around A2016 and the activation segment): Residues highlighted by the cryo-EM interaction pattern—especially A1950, E1920, D2017, and A2016—will emerge as dominant energetic contributors for type II ligands.
H3 (Energetics ↔ safety coupling): Inhibitors with stronger reliance on a narrow set of high-magnitude hotspots (high “energetic concentration”) will show less favorable predicted ADME-Tox (e.g., off-target liabilities), while scaffolds with more distributed stabilizing contributions will be predicted as safer and more optimizable for CNS use (testable via cross-model consensus ADME-Tox scoring).

3. Methods

3.1 Study design overview (deterministic, two-phase pipeline)

Phase 1 (Ligand-centric characterization; Weeks 1–6): Quantum/physicochemical descriptors + consensus AI ADME-Tox for the five inhibitors.
Phase 2 (Complex-centric analysis; Weeks 5–18): MD refinement of each cryo-EM complex, followed by quantum fragmentation (per-residue energetics) on representative snapshots.
Phase 3 (Integration + validation; Weeks 16–24): Hotspot validation (in silico alanine scanning / mutational perturbations), energetic clustering, and linking energetic patterns to ADME-Tox fingerprints. Deliver preregistered outputs and a preprint-ready manuscript draft.

3.2 Targets and inclusion criteria

Protein target: human LRRK2 kinase domain within the experimental cryo-EM complexes (using the deposited coordinates as templates).
Ligands (fixed panel; no cherry-picking): LRRK2-IN-1, GNE-7915 (type I); ponatinib, rebastinib, GZD-824 (type II).
Inclusion criteria: [i] Use the published co-structures of LRRK2 with each inhibitor (coordinate sets from the paper); [ii] If any loop/side chain is missing at the binding site, we will model missing atoms using standard refinement and validate with local stereochemistry checks; no changes to the inhibitor panel or target set after preregistration.

3.3 Procedures and data generation

Phase 1 — Ligand-centric quantum biochemistry + AI safety profiling

3.3.1 Ligand structure standardization

Generate 3D conformers for each inhibitor; assign protonation states at pH 7.4 (dominant microspecies) and relevant tautomers.
Minimize structures with a consistent protocol before quantum calculations.

3.3.2 DFT descriptor panel (quantum biochemistry core)
For each inhibitor, compute a standardized descriptor set analogous to the template proposal:

Electronic structure: HOMO/LUMO, gap, ionization potential proxies, chemical potential, hardness/softness, electrophilicity.
Charge distribution: RESP/CM5 partial charges (depending on software availability), dipole moment, molecular electrostatic potential (MEP) maps.
Solvation: implicit solvent single-point energies (e.g., SMD/PCM), polar surface area and related proxies.
These descriptors support mechanistic interpretation of why specific chemical motifs engage hotspots differently.

3.3.3 Consensus AI ADME-Tox fingerprint
We will compute a consensus “liability fingerprint” per inhibitor across multiple complementary predictors (ADMETlab/pkCSM/SwissADME-like and hERG/toxicity predictors; exact tools depend on institutional access but will be fixed before running). Outputs include: BBB penetration proxies, CYP interactions, hERG liability, hepatotoxicity proxies, genotox flags, and general drug-likeness alerts.
Primary output: a harmonized scorecard per inhibitor, plus model disagreement metrics.

Phase 2 — Complex-centric MD refinement and energetic mapping/signatures

3.3.4 System setup from cryo-EM templates

Start from the deposited LRRK2–inhibitor complexes (no redocking unless required for missing atoms).
Solvate in explicit water with physiological salt (0.15 M NaCl).
Assign protein force field (e.g., AMBER ff14SB/ff19SB equivalent) and ligand parameters (GAFF2/CGenFF-like with RESP/AM1-BCC charges as appropriate).
Minimize → heat → equilibrate with positional restraints released gradually.

3.3.5 Production MD

For each complex: 3 independent replicates × 100 ns (total 300 ns per ligand; 1.5 µs total for five ligands).
Record trajectories at fixed intervals (e.g., 10 ps).
QC endpoints: RMSD/RMSF stability, pocket contact persistence, hydrogen bond occupancy, water-mediated networks.

3.3.6 State/feature tracking tied to the cryo-EM mechanistic observations
We will explicitly quantify features implicated by the structural study, including:

Type II: αC helix shift, activation segment helix/loop occupancy, and contacts involving A1950/E1920/D2017/A2016.
s41421-023-00639-8
s41421-023-00639-8
Type I: similarity to ATP-bound active-like arrangement as a stability benchmark.
s41421-023-00639-8

3.3.7 Snapshot selection

Use the final 20 ns of each replicate after confirming stability.
Extract 100 evenly spaced frames per replicate (300 frames per inhibitor).
For each inhibitor, compute per-residue energetics on a representative subset (e.g., top 30 frames by pocket stability metrics), keeping runtime feasible.

3.3.8 MFCC-DFT (or equivalent fragmentation QM)

Fragment the protein around the binding site into residue-level fragments with capping scheme.
Compute ligand–residue interaction energies decomposed into electrostatics/dispersion/polarization surrogates (method-dependent), generating ΔE_residue maps per inhibitor.
Primary output: a per-residue energy table and “hotspot” ranking (top stabilizing and destabilizing residues).

Phase 3 — Hotspot validation and integration

3.3.9 In silico alanine scanning / mutation perturbations

Validate hotspots via alanine scanning on the binding site and (critically) A2016 perturbation as a targeted test for type II sensitivity. The structural data indicate A2016 directly contacts type II inhibitors but not ATP.
s41421-023-00639-8
Compare predicted affinity shifts to energetic rankings (consistency check).

3.3.10 Energetics × ADME-Tox integration

Create an integrated “physics-informed safety map”:
- Energetic concentration index (e.g., fraction of total stabilizing energy contained in top N residues).
- Correlate energetic concentration and key motif dependence with predicted ADME-Tox liabilities.

3.4 Ethical considerations and open science

This is a fully in silico reanalysis of public structural datasets; no human/animal subjects, thus no IRB/IACUC requirements.
All protocols (exact software versions, scripts, parameter files) will be deposited to an open repository (GitHub/Zenodo), including processed per-residue energy tables and ADME-Tox scorecards.

3.5 Timeline (6 months)

Month 1–1.5: Ligand DFT + ADME-Tox consensus (Phase 1).
Month 2–4: Replicated MD for all 5 complexes + QC (Phase 2).
Month 4–5: QM fragmentation/per-residue energetics + hotspot ranking (Phase 2b).
Month 5–6: Validation scans + integrated analysis + preprint draft (Phase 3).

4. Plan for statistical analysis

4.1 Primary endpoints

Hotspot overlap and class separation: similarity/dissimilarity of ΔE_residue vectors between inhibitors; primary comparison is type I vs type II.
Residue-level testing for A2016 and activation-segment network: confirm whether A2016 is significantly more stabilizing for type II than type I (directional hypothesis).
s41421-023-00639-8
Energetics–toxicity association: correlation between energetic concentration metrics and ADME-Tox consensus liabilities.

4.2 Statistical model and inference

Replicates: MD has n = 3 replicates per inhibitor (independent random seeds).
For each metric (e.g., ΔE_residue, H-bond occupancy), compute replicate means and 95% bootstrap CIs using frames as resampling units within replicate, then replicate-level aggregation.
Compare type I vs type II using a mixed-effects model (fixed effect: inhibitor class; random effect: inhibitor identity; nested random effect: replicate).
For residue-wise comparisons across many residues, control false discoveries with Benjamini–Hochberg FDR (q = 0.05).

4.3 Power / feasibility logic

This proposal is designed as a proof-of-concept within 6 months, prioritizing robust replicated dynamics and high-information QM decompositions on a bounded ligand panel. We will treat effect sizes as exploratory but will report [i] absolute and standardized differences (Cohen’s d) between classes, [ii] CIs, and [iii] full code to reproduce estimates.

4.4 Stopping rules / deviations

If a replicate fails to equilibrate (persistent drift, unstable ligand pose), it will be restarted once with a new seed; if failure repeats, that inhibitor will be reported as “u /removal of ligands after preregistration.

5. Budget

Lead Investigator / Postdoctoral Fellow: $6,000

Tasks: Preregistration package (ResearchHub), microstate and conformer design (biologically grounded), quantum-chemical workflow supervision and QC, primary interpretation and manuscript leadership, open-data packaging and governance, job preparation and execution support, data extraction and structuring, dataset QA and reproducibility checks.

Laboratory of Molecular Modeling of Biomolecules (M2Bio/UFRN): $3,000

Tasks: Full operational counterparty for non-personnel costs required to complete the Project, including: access to institutional computing environment and software stack required for quantum calculations; local storage infrastructure for raw and processed data; curated backup routines; research environment support—scientific oversight, lab operational support, and internal technical troubleshooting workflows.

High Performance Processing Center (NPAD/UFRN): In-Kind Support

Tasks:

HPC resource allocation and scheduling: Provide CPU node allocations suitable for high-throughput Opt/Freq/Raman/IR and TD-DFT batches; configure fair-share scheduling, queue policies, and job-array execution to support ensemble runs without bottlenecks.
Software environment provisioning: Provide and maintain the operational Gaussian environment on NPAD infrastructure (modules, libraries, scratch configuration); ensure consistent runtime environments across nodes for reproducibility (documented module versions, consistent binaries).
Job execution infrastructure: Provide high-throughput job submission mechanisms (arrays/batch scripts), monitoring, and automated notifications for failed jobs; provide scratch-space policies tuned for Gaussian I/O demands (fast scratch, cleanup policy, quota management).
Performance and reliability support: Troubleshoot computational failures attributable to environment issues (filesystem, node instability, scheduler interruptions); implement workload balancing recommendations to optimize throughput (e.g., memory-per-core constraints, scratch placement).
Data integrity, storage, and backup: Provide secure storage for raw Gaussian outputs and structured result tables; execute routine backups and integrity checks to prevent data loss across long runs.
Reproducibility and compliance support: Provide technical documentation confirming compute environment provenance (cluster type, OS baseline, scheduler, module versions) for inclusion in open-science metadata; support generation of checksums and archival bundles for Zenodo deposition.

Coordination for the Improvement of Higher Education Personnel (CAPES/Brazil): In-Kind Support

Tasks: This project will submit results to an open-access venue. Article processing charges (APCs) and/or open-access publication costs will be covered through CAPES/Brazil institutional agreements and CAPES-supported open-access funding arrangements.

6. References

Zhu H, Hixson P, Ma W, Sun J. Pharmacology of LRRK2 with type I and II kinase inhibitors revealed by cryo-EM. Cell Discovery. 2024;10:10.
Jennings D, et al. Preclinical and clinical evaluation of the LRRK2 inhibitor DNL201 for Parkinson’s disease. Science Translational Medicine. 2022;14:eabj2658.
Jennings D, et al. LRRK2 inhibition by BIIB122 in healthy participants and patients with Parkinson’s disease. Movement Disorders. 2023;38:386–398.
Miao Y, Feher VA, McCammon JA. Gaussian accelerated molecular dynamics: unconstrained enhanced sampling and free energy calculation. Journal of Chemical Theory and Computation. 2015;11:3584–3595.
Abraham MJ, et al. GROMACS: High performance molecular simulations through multi-level parallelism from laptops to supercomputers. SoftwareX. 2015;1–2:19–25.
Frisch MJ, et al. Gaussian 16, Revision C.01. Gaussian, Inc.; 2016.
de Lima Menezes, G., Sales Bezerra, K., Nobre Oliveira, J. I., Araújo, J. F., Galvão, D. S., da Silva, R. A., Saivish, M. V., & Fulco, U. L. (2024). Quantum mechanics insights into melatonin and analogs binding to melatonin MT1 and MT2 receptors. Scientific Reports, 14, 10922.
Lima Neto, J. X., Sales Bezerra, K., Barbosa, E. D., Araujo, R. L., Galvão, D. S., Lyra, M. L., Nobre Oliveira, J. I., Akash, S., Bin Jardan, Y. A., Nafidi, H.-A., Bourhia, M., & Fulco, U. L. (2024). Investigation of protein-protein interactions and hotspot region on the NSP7-NSP8 binding site in NSP12 of SARS-CoV-2. Frontiers in Molecular Biosciences, 10, 1325588.
Esmaile, S. C., Sales Bezerra, K., de Oliveira Campos, D. M., da Silva, M. K., Lima Neto, J. X., Manzoni, V., Fulco, U. L., & Nobre Oliveira, J. I. (2021). Quantum binding energy features of the drug olmesartan bound to angiotensin type-1 receptors in the therapeutics of stroke. New Journal of Chemistry, 45, 19487–19496.
Lima Neto, J. X., de Oliveira Campos, D. M., Nobre Oliveira, J. I., Galvão, D. S., & Fulco, U. L. (2023). In Silico Evaluation of the Binding Energies of Androgen Receptor Agonists in Wild-Type and Mutational Models. The Journal of Physical Chemistry B, 127(22), 5005–5017.
Costa, A. H. L., Sales Bezerra, K., Lima Neto, J. X., Nobre Oliveira, J. I., Galvão, D. S., & Fulco, U. L. (2023). Deciphering Interactions between Potential Inhibitors and the Plasmodium falciparum DHODH Enzyme: A Computational Perspective. The Journal of Physical Chemistry B, 127(44).

Energetic and Safety Signatures of Type I vs Type II LRRK2 Kinase Inhibitors: A Quantum Biochemistry–Guided Reanalysis of Cryo-EM Structures for Parkinson’s Therapeutic Design

Energetic and Safety Signatures of Type I vs Type II LRRK2 Kinase Inhibitors: A Quantum Biochemistry–Guided Reanalysis of Cryo-EM Structures for Parkinson’s Therapeutic Design

Funding Progress

Related Funding

Request for Proposals - Neuroscience

Nonprofit Recipient

Topics

DOI