AlphaCert

Automated Quality-Control, pLDDT/PAE Assessment, and Stereochemical Certification for Predicted Protein Structures (AlphaFold2/3, ColabFold, ESMFold).

Overview

AlphaCert is an open-source scientific toolkit designed to systematically audit, validate, and certify the reliability of 3D protein structures predicted by AlphaFold2, AlphaFold-Multimer, AlphaFold3, ColabFold, and ESMFold.

Rather than blindly relying on predicted models for molecular docking, molecular dynamics, or drug discovery campaigns, alphacert performs an automated multi-criteria structural audit in a single command:

• 📊 Residue pLDDT Confidence Profile & Classification:
  • Categorizes residues into the 4 canonical AlphaFold bands:
    • 🔵 Very High Confidence ($> 90$): Suitable for atomic side-chain modeling and docking.
    • 🔷 Confident ($70 - 90$): Reliable backbone topology.
    • 🟡 Low Confidence ($50 - 70$): Loop regions or flexible segments.
    • 🟠 Very Low Confidence ($< 50$): Intrinsically Disordered Regions (IDRs).
• 🧩 2D Predicted Aligned Error (PAE) Matrix Audit:
  • Identifies rigid domains by clustering the PAE matrix with Croll's greedy-modularity method (the method used by ChimeraX and AlphaFold DB; cutoff 5 Å). Disordered clusters with mean pLDDT < 70 are excluded.
  • Evaluates multimer interfaces with ipTM as the primary criterion (> 0.8 confident, 0.6–0.8 grey zone, < 0.6 likely failed). The symmetrised inter-chain PAE is used when no ipTM is available.
• 📐 Stereochemical & Geometry Quality:
  • Heavy-atom clash score: non-bonded pairs (more than 3 covalent bonds apart) with $d_{ij} < r_{\text{vdw}, i} + r_{\text{vdw}, j} - 0.40$ Å, per 1000 heavy atoms. This is a proxy without hydrogens, not the MolProbity clashscore.
  • Ramachandran dihedral distribution ($\phi, \psi$ Favored $\ge 88%$, Outliers $< 3%$) using approximate rectangular regions, not the MolProbity Top8000 contours. For publication-grade figures, use MolProbity or PHENIX.
  • Non-proline cis-peptide bond identification.
• 🎯 Target Application Readiness Certification:
  • 💊 DOCKING_READY: Verifies that active/binding site residues have $\text{pLDDT} \ge 80$.
  • 🌊 MD_READY: Verifies that disordered terminal loops do not compromise solvated simulation stability.
  • 🤝 INTERFACE_VALID: Distinguishes authentic protein-protein interfaces from non-binding artifacts.
• 📑 Publication Deliverables:
  • Interactive self-contained report.html dashboard.
  • Publication vector plots (pLDDT sequence profile, 2D PAE error heatmap, Ramachandran scatter plot) in SVG, PDF, PNG (300 DPI).
  • Ready-to-compile LaTeX summary tables (.tex).
  • Draft Methods text snippet and BibTeX citation (citation.bib).

   AlphaFold PDB / mmCIF + PAE JSON
                  │
                  ▼
  ┌───────────────────────────────────────────────────────────┐
  │                         AlphaCert                         │
  │  ├── Residue-level pLDDT Band Categorization & IDR Audit  │
  │  ├── 2D Predicted Aligned Error (PAE) Domain Decomposition│
  │  ├── Steric Clashscore & Ramachandran Geometry Check      │
  │  └── Downstream Certification (Docking-Ready / MD-Ready)  │
  └───────────────────────────────────────────────────────────┘
                  │
                  ▼
  ┌───────────────────────────────────────────────────────────┐
  │                   Publication Deliverables                │
  │  ├── report.html (Interactive Dashboard & Badges)         │
  │  ├── alphacert_plddt_profile.pdf/svg/png                  │
  │  ├── alphacert_pae_matrix.pdf/svg/png                     │
  │  ├── alphacert_ramachandran_plot.pdf/svg/png              │
  │  ├── alphacert_summary_table.tex / .csv                   │
  │  ├── methods_snippet.txt (Ready for Manuscript)           │
  │  └── citation.bib (BibTeX Reference)                      │
  └───────────────────────────────────────────────────────────┘

Installation

From PyPI

Note: PyPI release pending. Until then, install from the tagged GitHub release:

pip install "git+https://github.com/sircalch/alphacert@v1.0.0"

From Source

git clone https://github.com/sircalch/alphacert.git
cd alphacert
pip install -e .[dev]

Quickstart (CLI)

1. Run Benchmark Demo (Instant AlphaFold Kinase Model Audit)

alphacert demo -o my_protein_audit/

Open my_protein_audit/report.html in any browser to inspect the interactive report!

2. Assess Predicted PDB Structure

alphacert assess -i predicted_structure.pdb --pae predicted_pae.json -o struct_report/

3. Certify Binding Pocket for Molecular Docking

alphacert assess -i model.pdb --pocket "45,46,47,78,120,121" -o docking_check/

Python API Usage

from alphacert import assess_alphafold_quality
from alphacert.parsers import parse_protein_structure, parse_pae_matrix
from alphacert.reporters import generate_alphacert_figures, generate_alphacert_manuscript_assets, generate_alphacert_html_report

# 1. Parse structure and PAE matrix
struct = parse_protein_structure("alphafold_model.pdb")
pae = parse_pae_matrix("alphafold_pae.json")

# 2. Assess structure quality
report = assess_alphafold_quality(
    metadata={"name": "Target Kinase", "engine": "AlphaFold2"},
    plddt_values=struct["per_residue_plddt"],
    pae_matrix=pae["pae_matrix"],
    coordinates=struct["coordinates"],
    atom_elements=struct["atom_elements"],
    backbone_atoms=struct["backbone_atoms"],
    binding_site_indices=[45, 46, 47, 78, 120]
)

print(f"Overall Certification: {report.overall_status}")
print(f"Mean pLDDT: {report.plddt_result.mean_plddt:.1f}")
print(f"Docking Readiness: {report.docking_readiness['status']}")

# 3. Export publication deliverables
generate_alphacert_figures(report, "output_dir/")
generate_alphacert_manuscript_assets(report, "output_dir/")
generate_alphacert_html_report(report, "output_dir/report.html")

Citation

If you use AlphaCert in your research, please cite:

@software{monreal2026alphacert,
  author = {Monreal-Hern{\'a}ndez, Andre},
  title = {{AlphaCert: An Open-Source Toolkit for Quality-Control, pLDDT/PAE Assessment, and Stereochemical Certification of Predicted Protein Structures}},
  year = {2026},
  version = {1.0.0},
  publisher = {Zenodo},
  url = {https://github.com/sircalch/alphacert}
}

License

This project is licensed under the MIT License - see the LICENSE file for details.