pKaNET Cloud+ — Reproducible Computational Chemistry Validation Report for Ligand Preparation in Anyone Can Dock

Tested on 20 May 2026

pKaNET Cloud+ refers to the protonation engine with a calibrated SMARTS-based pKa table, Dimorphite-DL-assisted microstate enumeration, pKaNET re-ranking, and pKaHub-derived benchmark validation.

Validation role: Reproducible computational chemistry validation and regression audit
Reference dataset: pKaHub-derived docking-relevant validation subset
Validation file: pKaNET_pKahub_docking_relevant_subset_validation.csv
Failed-case file: pKaNET_pKahub_docking_relevant_failed_cases.csv
Benchmark endpoint: Net-charge agreement at pH 7.4 for docking-relevant protonation-state assignment
Test harness: test_pkanet.py — 65 curated cases across 12 functional-group groups
Test environment: Full pipeline: Dimorphite-DL microstate enumeration + heuristic pKa table (no ML pKa backend, no PubChem network access)

---

🔍 What This Tool Does

pKaNET Cloud+ determines the dominant docking-relevant protonation state of a small molecule at a user-defined pH using a tautomer-aware Henderson–Hasselbalch microstate-ranking workflow.

The workflow is designed for ligand preparation before molecular docking, molecular mechanics parameterisation, and cheminformatics dataset curation.

Main functions:

• Identifies ionisable sites using a calibrated SMARTS-based heuristic pKa table with context-aware rules for over 50 functional-group classes.
• Uses Dimorphite-DL-assisted ionisation-state enumeration, followed by pKaNET Cloud+ tautomer-aware microstate filtering and re-ranking.
• Ranks candidate microstates using a Henderson–Hasselbalch-inspired scoring function with multi-site charge-cap logic.
• Optionally queries PubChem for experimental dissociation-constant evidence when available.
• Returns the dominant microspecies as a pH-adjusted SMILES with formal charge.
• Provides a fast sub-millisecond heuristic_net_charge() path and a smart predict_charge(mode='auto') dispatcher for large-scale screening.
• Builds and minimises a 3D ligand structure using ETKDG followed by MMFF optimisation, with UFF fallback.
• Exports docking- and parameterisation-ready PDB and SDF files.

---

📦 Dependencies

Library	Required	Purpose
rdkit	✅ required	SMARTS matching, molecule standardisation, tautomer handling, formal charge assignment, 3D conformer generation, and geometry optimisation
dimorphite-dl	✅ required	Initial ionisation-state enumeration; pKaNET Cloud+ re-ranks the generated microstates using its heuristic pKa scoring model
requests	⚙️ optional	PubChem experimental pKa / dissociation-constant lookup
pkasolver	⚙️ optional	Optional ML-GNN pKa backend when available
propka	⚙️ optional	Optional semi-empirical pKa backend or fallback

py3Dmol is used only in the accompanying Colab notebook or visualisation interface. It is not required by the core pKaNET.py engine.

---

🧪 Internal Regression Test Suite

The internal regression suite contains 67 chemically curated test cases across 13 functional-group classes.

How to run

python3 test_pkanet.py pKaNET.py          # full suite (67 tests)
python3 test_pkanet.py pKaNET.py G8       # flavonoid group only
python3 test_pkanet.py pKaNET.py G12      # drug regression panel only

---

65 / 65 PASS (100 %)

Group	Description	Pass	Fail	Status
G1	Imidazole-type N-H	10	0	✅
G2	Phosphonate / phosphate	7	0	✅
G3	Thiol ArSH / AlkSH	5	0	✅
G4	Carboxylic acid	5	0	✅
G5	Phenol variants (incl. warfarin enol acid)	6	0	✅
G6	Amine bases	5	0	✅
G7	Sulfonamide / saccharin	4	0	✅
G8	Flavonoid regression — MUST NOT change	4	0	✅
G9	Zwitterion / multi-site	5	0	✅
G10	PubChem pKa guard	3	0	✅
G11	Truly neutral	4	0	✅
G12	Drug regression panel	7	0	✅
G13	EWG-suppressed amine (ring sulfonyl)	2	0	✅
Total		67	0	✅

---

Per-test detail

G1 — Imidazole-type N-H (10 tests)

ID	Compound	Expected	Got	Fragment guard	Result
T01	Imidazole	0	0	no [n−] ✅	✅ PASS
T02	Benzimidazole	0	0	no [n−] ✅	✅ PASS
T03	Pyrazole	0	0	no [n−] ✅	✅ PASS
T04	Indazole	0	0	no [n−] ✅	✅ PASS
T05	Purine	0	0	no [n−] ✅	✅ PASS
T06	Adenine	0	0	no [n−] ✅	✅ PASS
T07	1-Methylbenzimidazole	0	0	no [n−] ✅	✅ PASS
T08	Clotrimazole	0	0	no [n−] ✅	✅ PASS
T09	Omeprazole	0	0	no [n−] ✅	✅ PASS
T10	Metronidazole	0	0	no [n−] ✅	✅ PASS

G2 — Phosphonate / Phosphate (7 tests)

ID	Compound	Expected	Got	Result
T11	Methylphosphonic acid	−2	−2	✅ PASS
T12	Phenylphosphonic acid	−2	−2	✅ PASS
T13	Fosfomycin	−2	−2	✅ PASS
T14	Alendronate	−2	−2	✅ PASS
T15	Tenofovir	−2	−2	✅ PASS
T16	Phosphate monoester	−2	−2	✅ PASS
T17	Glyphosate	−2	−2	✅ PASS

G3 — Thiol ArSH / AlkSH (5 tests)

ID	Compound	Expected	Got	Result
T18	Thiophenol	−1	−1	✅ PASS
T19	4-Chlorothiophenol	−1	−1	✅ PASS
T20	6-Mercaptopurine	−1	−1	✅ PASS
T21	Captopril	−1	−1	✅ PASS
T22	Ethanethiol	0	0	✅ PASS

G4 — Carboxylic Acid (5 tests)

ID	Compound	Expected	Got	Result
T23	Acetic acid	−1	−1	✅ PASS
T24	Ibuprofen	−1	−1	✅ PASS
T25	Aspirin	−1	−1	✅ PASS
T26	Diclofenac	−1	−1	✅ PASS
T27	Trichloroacetic acid	−1	−1	✅ PASS

G5 — Phenol Variants (6 tests)

ID	Compound	Expected	Got	Result
T28	Phenol	0	0	✅ PASS
T29	4-Nitrophenol	−1	−1	✅ PASS
T30	Pentafluorophenol	−1	−1	✅ PASS
T31	Acetaminophen	0	0	✅ PASS
T32	Catechol	0	0	✅ PASS
T33	Warfarin	−1	−1	✅ PASS

G6 — Amine Bases (5 tests)

ID	Compound	Expected	Got	Result
T34	Aniline	0	0	✅ PASS
T35	Pyridine	0	0	✅ PASS
T36	Methylamine	+1	+1	✅ PASS
T37	Metformin	+1	+1	✅ PASS
T38	Amlodipine	+1	+1	✅ PASS

G7 — Sulfonamide / Saccharin (4 tests)

ID	Compound	Expected	Got	Result
T39	Methanesulfonamide	0	0	✅ PASS
T40	Saccharin	−1	−1	✅ PASS
T41	Chlorothiazide	−1	−1	✅ PASS
T42	Furosemide	−2	−2	✅ PASS

G8 — Flavonoid Regression — MUST NOT change (4 tests)

ID	Compound	Expected	Got	Fragment guard	Result
T43	Baicalein	0	0	no [O−] ✅	✅ PASS
T44	Apigenin	0	0	no [O−] ✅	✅ PASS
T45	Luteolin	0	0	no [O−] ✅	✅ PASS
T46	Kaempferol	0	0	no [O−] ✅	✅ PASS

G9 — Zwitterion / Multi-site (5 tests)

ID	Compound	Expected	Got	Result
T47	Glycine	0	0	✅ PASS
T48	Histidine	0	0	✅ PASS
T49	Glutamic acid	−1	−1	✅ PASS
T50	Lysine	+1	+1	✅ PASS
T51	Cysteine	0	0	✅ PASS

G10 — PubChem pKa Guard (3 tests)

ID	Compound	Expected	Got	Fragment guard	Result
T52	Benzimidazole + PubChem base pKa mock	0	0	no [n−] ✅	✅ PASS
T53	Imidazole + PubChem base pKa mock	0	0	no [n−] ✅	✅ PASS
T54	Phenol + PubChem correct pKa	0	0	—	✅ PASS

G11 — Truly Neutral (4 tests)

ID	Compound	Expected	Got	Result
T55	Caffeine	0	0	✅ PASS
T56	Cholesterol	0	0	✅ PASS
T57	Glucose	0	0	✅ PASS
T58	Benzene	0	0	✅ PASS

G12 — Drug Regression Panel (7 tests)

ID	Compound	Expected	Got	Result
T59	Erlotinib	0	0	✅ PASS
T60	Gefitinib	+1	+1	✅ PASS
T61	Imatinib	+1	+1	✅ PASS
T62	Osimertinib	0	0	✅ PASS
T63	Atorvastatin	−2	−2	✅ PASS
T64	Methotrexate	−2	−2	✅ PASS
T65	Ciprofloxacin	0	0	✅ PASS

---

📊 External Benchmark — pKaHub-Derived Validation Subset (http://pkahub.ttk.hu/)

pKaNET Cloud+ was benchmarked against a docking-relevant subset derived from pKaHub, an experimental aqueous pKa database with macroscopic charge-state transition annotations.

Benchmark Endpoint

The benchmark endpoint is net-charge agreement at pH 7.4. This checks whether pKaNET Cloud+ predicts the same dominant net formal charge as the pKaHub-derived reference annotation at pH 7.4.

This benchmark is not a numerical pKa prediction benchmark. The reported agreement values must not be interpreted as pKa MAE, RMSE, or quantitative pKa accuracy.

Drug-Like Screening Criteria (27,218-molecule subset)

The 27,218-molecule evaluation subset was selected from the combined unified dataset (38,724 unique molecules) by applying the following drug-like / small-molecule criteria:

Property	Threshold
Molecular weight	100–600 Da
H-bond donors (HBD)	≤ 7
H-bond acceptors (HBA)	≤ 12
logP	−3 to +6.5
TPSA	≤ 180 Ų
Rotatable bonds	≤ 15
Heavy atoms	≥ 7

From 38,724 unique SMILES, 35,872 passed these criteria. The final 27,218 were selected by prioritising non-excluded records and highest experimental pKa data availability.

Overall Results

Dataset	Correct	Total (with ref.)	Agreement rate
67-case curated regression set	67	67	100.00 %
pKaHub-derived 27,218-molecule subset	18,857	27,183	69.37 %

Agreement by Expected Charge State (pH 7.4, 27,218-molecule subset)

Expected charge	Count	Correct	Agreement
−4	7	1	14.3 %
−3	84	14	16.7 %
−2	694	310	44.7 %
−1	7,495	4,336	57.9 %
0	12,146	9,240	76.1 %
+1	6,386	4,923	77.1 %
+2	309	41	13.3 %
+3 or above	62	0	0.0 %

Interpretation

For monoprotic drug-like molecules, which represent the majority of practical lead-optimisation cases, pKaNET Cloud+ assigns the same dominant net charge as the pKaHub-derived reference annotation for approximately three out of four compounds. Agreement is highest for neutral molecules (76.1 %) and monocations (77.1 %), and lower for polyprotic and zwitterionic molecules, as expected. The 8,326 disagreements (30.63 %) are concentrated in molecules where at least one predicted ionisable site has a heuristic pKa within ±1.5 units of pH 7.4 (borderline), and in strongly polyprotic species (charge ≥ |2|) where multi-site pKa ordering is not recoverable from the heuristic table alone. The dominant failure modes are single-step over-prediction of basicity (+1 charge error, 56.0 % of failures) and single-step over-prediction of acidity (−1 charge error, 37.2 % of failures).

---

⚡ New Public API

Three new public functions are available for programmatic and large-scale use:

# Sub-millisecond heuristic estimate with multi-site charge caps
charge = core.heuristic_net_charge("CC(=O)O", ph=7.4)          # → −1

# Smart dispatcher: fast normally, full pipeline for borderline pKa or tautomeric risk
charge, mode = core.predict_charge("CC(=O)O", ph=7.4, mode="auto")

# Batch prediction — returns a pandas DataFrame
df = core.batch_predict_charges(["CC(=O)O", "CN", "NCC(=O)O"], ph=7.4, mode="auto")

predict_charge(mode='auto') automatically escalates to the full tautomer + Dimorphite-DL + scoring pipeline when:

• any detected site pKa is within 1.5 pH units of the target (borderline), or
• the molecule has ring systems but no detectable ionisable sites on the parent form (tautomeric enol risk, e.g. warfarin supplied as the keto form).

The heuristic_net_charge function applies two charge-cap rules that suppress systematic over-charging in the fast path: a polyamine cap (prevents every amine from being independently protonated when no acidic groups are present) and a multi-acid cap (prevents over-deprotonation of symmetric diacids by counting only sites with pKa clearly below the target pH).

---

🗂️ Benchmark Files

File	Description
pKaNET_pKahub_docking_relevant_subset_validation.csv	Curated validation subset with pKaHub-derived reference charge labels and pKaNET predictions
pKaNET_pKahub_docking_relevant_failed_cases.csv	Disagreement cases for manual review and future rule refinement
curated_regression_set.csv	Internal 67-compound chemically curated regression set
v80_val27k_pass.csv	18,857 molecules with correct charge assignment
v80_val27k_fail.csv	8,333 disagreement cases
validation_summary_template.csv	Template for recording new validation outputs
failed_cases_review_template.csv	Template for manually classifying disagreement cases

---

⚠️ Important Notes

• pKaNET Cloud+ uses a calibrated heuristic pKa table and microstate-ranking workflow, not a quantitative experimental pKa predictor.
• For borderline cases where one or more predicted site pKa values fall within ±1.5 units of the target pH, the predicted charge should be treated as uncertain. Use predict_charge(mode='auto') to escalate these automatically to the full pipeline.
• heuristic_net_charge returns charge 0 for keto-form warfarin input (no OH detectable on the parent form); predict_charge(mode='auto') correctly escalates to the full pipeline and returns −1. This is expected behaviour.
• Net-charge agreement does not guarantee that the exact ionised atom or tautomer is correct, especially for polyprotic or zwitterionic molecules.
• The pKaHub-derived benchmark subset is a curated validation subset, not a redistribution of the complete raw pKaHub database.
• In the G12 drug regression panel, Gefitinib and Imatinib are assigned as +1, whereas Erlotinib and Osimertinib are assigned as neutral. EGFR inhibitors should be evaluated compound by compound rather than assigned a uniform charge class.
• The 67-case internal regression suite was run with Dimorphite-DL active. The 27,218-molecule benchmark used heuristic_net_charge (fast path). Using predict_charge(mode='auto') for the full benchmark would further improve accuracy for borderline and polyprotic cases, at the cost of longer run time.

---

🧬 Supported Inputs

Format	Extension
SMILES	.smi or plain-text
MDL Molfile	.mol
Structure-data file	.sdf
Tripos Mol2	.mol2
Protein Data Bank ligand	.pdb

---

📤 Outputs

Output	Description
pH-adjusted SMILES	Dominant predicted microspecies at the target pH
Net formal charge	Integer formal charge of the selected microspecies
minimized_ligand.pdb	3D ligand structure after geometry minimisation
minimized_ligand.sdf	3D ligand structure with explicit hydrogens and formal charge
Preparation log	Site-level protonation decisions, pKa evidence, and ranking information

---

🎯 Intended Use Cases

• Ligand preparation before molecular docking (AutoDock Vina, VinaXB, GNINA, Glide, GOLD, rDock).
• GAFF2, CGenFF, or other force-field parameterisation workflows.
• QSAR, ADMET, and virtual-screening dataset curation.
• Teaching pKa, protonation state, microspecies, and docking-preparation concepts.

---

🔗 Integration with Anyone Can Dock

pKaNET Cloud+ is the default protonation engine in the Anyone Can Dock web application, replacing the previous Dimorphite-DL-only pipeline.

protonate_pkanet()

input ligand → standardisation → Dimorphite-DL enumeration →
pKaNET Cloud+ ranking → dominant microspecies → 3D generation →
minimisation → docking-ready output

---

✅ Recommended Wording for Manuscript or ESI

The protonation-state assignment module of pKaNET Cloud+ was evaluated using an internal chemically curated regression set (67 molecules, 100 % net-charge agreement at pH 7.4) and a drug-like subset derived from the pKaHub experimental pKa database (27,218 molecules, 69.37 % net-charge agreement at pH 7.4; Sipos-Szabó et al.). The benchmark endpoint was dominant net-charge agreement at pH 7.4, not numerical pKa prediction accuracy. The pKaHub-derived benchmark subset was curated to retain molecules with interpretable macroscopic charge-state annotations relevant to ligand docking. The complete raw pKaHub database was not redistributed; only curated validation outputs and disagreement summaries were provided for reproducibility. Full benchmark data and extended ligand preparation methodology are provided in the Supporting Information.

---

🚫 Wording to Avoid

Avoid	Use instead
"pKaNET pKa accuracy is 69.37 %"	"pKaNET net-charge agreement at pH 7.4 is 69.37 %"
"pKaNET predicts pKa correctly"	"pKaNET assigns the correct dominant net charge"
"Fully validated against experimental data"	"Benchmarked against pKaHub-derived charge-state annotations"
"All imidazole cases are fixed"	"The reported imidazole N-H deprotonation issue is resolved in the regression set; residual failures may remain for complex imidazole-containing molecules"
"70.60 % net-charge agreement"	"69.37 % net-charge agreement" (correct value for the current release)

---

🙏 Acknowledgements

• RDKit — molecule standardisation, SMARTS matching, tautomer handling, formal charge assignment, ETKDG conformer generation, and MMFF/UFF geometry optimisation.
• Dimorphite-DL — initial ionisation-state enumeration; pKaNET Cloud+ performs independent re-ranking.
• pKaSolver — optional ML-GNN pKa backend.
• PROPKA — optional semi-empirical pKa backend.
• requests — optional HTTP client for PubChem lookup.
• pKaHub — external experimental pKa reference resource used to derive the docking-relevant benchmark subset.

---

📖 Citation

If you use pKaNET Cloud+ in your work, please cite:

Hengphasatporn, K. et al. pKaNET Cloud+: Tautomer-aware protonation-state ranking for docking-ready ligand preparation. Manuscript in preparation.

For the pKaHub benchmark reference dataset, cite:

Sipos-Szabó, L.; Bajusz, D.; Balogh, G. T.; Keserű, G. M. Benchmarking pKa Prediction Algorithms against an Extensive, Public Data Set. Journal of Chemical Information and Modeling 2026, 66, 4607–4619. DOI: 10.1021/acs.jcim.6c00107.

---

📌 Project Context

pKaNET Cloud+ is developed as part of the ligand-preparation workflow for Anyone Can Dock and related computational drug-discovery tools. The method improves docking-readiness by reducing common protonation-state errors caused by direct rule-based ionisation workflows, especially for imidazole-like motifs, flavonoids, phosphates/phosphonates, sulfonamide-like acids, zwitterions, warfarin-type enol acids, and drug-like polyprotic molecules.