Cambridge MRC-LMB & IMB Research Framework • v0.4.5

Structural Identification of Anomalous Carboxyl Protonation & Low-Barrier Hydrogen Bonds

A rigorous crystallographic validation platform engineered to separate genuine shared-proton Low-Barrier Hydrogen Bonds (LBHBs) from refinement artifacts, thermal disorder, and unmodeled cation complexes across the Protein Data Bank (PDB).

Suleiman M Hajizadeh Institute of Molecular Biology, Ministry of Science and Education of Azerbaijan • Western Caspian University
Garib N Murshudov MRC Laboratory of Molecular Biology (LMB), Cambridge Biomedical Campus, United Kingdom
7,230
High-Res PDB Structures Screened
≤ 1.20 Å Resolution Limit
6,759
Total Carboxyl Pairs Mined
d(O···O) ≤ 3.20 Å Contacts
≤ 2.55 Å
Ultra-Short LBHB Boundary
Shared Proton • Strong pKa Shift
67
Golden Validated Benchmarks
Passed All 4 Murshudov Tiers
SCIENTIFIC VISION

The Fundamental Crystallographic Challenge

Why atomic distance alone is deceptive without rigorous electron density and thermal displacement validation.

The False Positive Hazard

In standard crystallographic refinement, side-chain carboxylates (Asp, Glu) in highly flexible loops can be artificially fitted into ambiguous, low-density features. A measured distance of d ≤ 2.50 Å often represents a modeling artifact or unmodeled metal bridge rather than a true hydrogen bond.

"If there is no electron density for atoms, they are mismodeled. If atoms are mismodeled, the distance between them is completely meaningless." — Prof. Garib Murshudov (MRC-LMB Cambridge)

The Carbox 4-Stage Precision Solution

Rather than blindly accepting PDB coordinates, Carbox applies a multi-tier decision tree combining Tukey Interquartile Range (IQR) outlier detection on isotropic B-factors, Fisher z-transformed real-space correlation coefficients (RSCC), empirical PROPKA pKa shift evaluation, and 70-metal coordination screening.

  • Eliminates thermal disorder noise (α = 1.5 & 3.0 Tukey whiskers)
  • Discards unmodeled Zn²⁺, Mg²⁺, Ca²⁺ deprotonated salts
  • Confirms genuine protonation shifts (ΔpKa ≥ +2.0)
METHODOLOGY

Prof. Garib Murshudov's 4-Stage Protocol

From raw high-resolution diffraction coordinates to verified biochemical protonation states.

STAGE 1

Euclidean Contact Mining & Solvent Burial

All atomic-resolution macromolecular crystal structures (≤ 1.20 Å) are indexed. The algorithm extracts side-chain carboxyl oxygens (Asp OD1/OD2, Glu OE1/OE2) and computes all inter-residue Euclidean pairwise distances:

$$d(O_i, O_j) = \sqrt{(x_i - x_j)^2 + (y_i - y_j)^2 + (z_i - z_j)^2} \le 3.20\,\text{Å}$$

The centroid of each pair is computed and cross-referenced with all crystallographic water molecules. Pairs with d(water) > 3.0 Å are tagged as Buried (ε ≈ 2–4, high dielectric penalty favoring protonation).

Stage 1 Screening Thresholds

  • Crystallographic Resolution≤ 1.20 Å (Tier 1)
  • Carboxyl Contact Cutoffd ≤ 3.20 Å
  • Ultra-Short LBHB Cutoffd ≤ 2.55 Å
  • Solvent Exclusion Cutoffd(water) > 3.00 Å
STAGE 2

Crystallographic Precision & Noise Pruning

As instructed by Prof. Garib Murshudov, pairs are subjected to a rigorous dual-prong validation protocol:

A. Non-Parametric Tukey IQR Filter
$$B_{\text{cutoff}} = Q_3 + \alpha \cdot \text{IQR} \quad (\text{IQR} = Q_3 - Q_1, \; \alpha \in \{1.5, 3.0\})$$

If either oxygen has $B > B_{\text{cutoff}}$ or $B_z = (B - \mu_B)/\sigma_B > 3.0$, the contact is classified as thermal disorder noise and rejected.

B. Real-Space Correlation with Fisher Z-Transform
$$z = \frac{1}{2} \ln \left(\frac{1 + \text{CC}'}{1 - \text{CC}'}\right), \quad \text{CC}' = \min(0.9999, \text{CC})$$

Atoms with standardized $Z_{\text{CC}} < -3.0$ exhibit poor density agreement and are pruned.

Stage 2 Filter Results

Pruned Disordered Loops:412 sites
Pruned Steric Clashes:287 sites
Pruned Metal Cations:49 sites
Validated Signal:67 sites
STAGE 3

Biochemical Protonation Modeling (PROPKA)

Isolated carboxylates in neutral water have standard pKa values of 3.8 (Asp) and 4.5 (Glu). Bringing two negative carboxylates to within 2.50 Å requires a tremendous electrostatic penalty that is only stabilized by a shared bridging proton.

$$\Delta \text{p}K_a = \text{p}K_a^{\text{local}} - \text{p}K_a^{\text{model}} \ge +2.0 \quad \Longrightarrow \quad \text{p}K_a \ge 6.0\text{--}8.0$$

Using PROPKA 3 integration, Carbox verifies that candidate LBHB carboxyl pairs exhibit dramatic upward pKa shifts, confirming that one residue functions as a neutral hydrogen donor at physiological pH.

Protonation Landscape

  • Aqueous Asp pKa3.80
  • Aqueous Glu pKa4.50
  • Buried LBHB Asp pKa6.50 – 8.20
  • Protonation ConfidenceStrong (>95%)
STAGE 4

10–20 Golden Benchmarks & Literature Validation

Before global batch scaling, the methodology is verified manually against 10–20 universally celebrated, experimentally proven catalytic and structural LBHBs:

  • 1KZK: HIV-1 Retroviral Protease catalytic dyad (Asp25···Asp25')
  • 1OEW: Endothiapepsin sub-angstrom aspartic dyad (0.90 Å)
  • 1FKN: BACE-1 Alzheimer beta-secretase dyad (Asp32···Asp228)
  • 2DSX: Rubredoxin sub-atomic buried core (0.68 Å, Asp35···Asp47)
  • 4ZM7: Aminopeptidase proton-relay network (0.70 Å, Asp149···Glu162)

Publication Rigor

Every identified golden benchmark is verified against primary crystallographic papers (Science, Nature, Structure) and registered with full DOI citations.

INTERACTIVE 3D EXPLORATION

Live Benchmark Molecular Showcase

Inspect real atomic structures, side-chain orientations, and LBHB distance vectors rendered directly via WebGL.

Select Case Study:

HIV-1 Protease Catalytic Dyad

PDB Accession: 1KZK
Crystallographic Resolution: 1.09 Å
Carboxyl Dyad: Asp25 ··· Asp25'
Measured Distance: 2.45 Å (Ultra-Short LBHB)
Solvent Environment: Buried (d_water > 3.0 Å)
Quality Status: Golden Reliable (Murshudov Tier A)

Symmetrical homodimer interface where two invariant Asp25 residues share a central proton with elevated pKa (>6.5) to perform general acid-base catalysis of viral polyprotein cleavage.

Left-Click: Rotate Scroll: Zoom Right-Click: Pan
VALIDATED DATASET

The 10–20 Golden Benchmark Database

Filter and inspect curated high-confidence LBHB sites verified against crystallographic quality metrics.

PDB Resolution Residue Pair d(O···O) d(Water) Max B-factor Tukey Limit Verdict / Role
OPEN SCIENCE & CITATION

Cite the Carbox Platform

Carbox is free and open-source scientific software. Please cite our research paper when utilizing these benchmarks or algorithms.

BibTeX Reference 
@article{HajizadehMurshudov2026,
  author    = {Hajizadeh, Suleiman M and Murshudov, Garib N},
  title     = {Structural Identification of Anomalous Carboxyl Protonation States,
               Low-Barrier Hydrogen Bonds, and Microenvironment Dielectric Shielding
               in High-Resolution Protein Crystal Structures},
  journal   = {Carbox & CarboxMol Macromolecular Crystallography Suite},
  year      = {2026},
  volume    = {0},
  number    = {4},
  pages     = {1--35},
  publisher = {Project Carbox},
  url       = {https://project-carbox.github.io/}
}

Research Artifacts & Code

Access reproducible pipelines, Jupyter/Colab notebooks, and pre-compiled benchmark datasets:

Google Colab Notebook Interactive execution of Murshudov 4-stage pipeline
Core Python Package github.com/project-carbox/carbox (CLI, filters, REST API)
CarboxMol 3D Workbench Desktop & WebGL electron density streaming suite