pkanet-cloud 0.9.0

Tautomer-aware microstate ranking · HH scoring · PubChem pKa evidence · pH-adjusted 3D structures

pKaNET Cloud+

Tautomer-aware microstate ranking · Henderson–Hasselbalch scoring · PubChem pKa evidence · pH-adjusted 3D structures

Part of the DFDD Project.

---

Features

• Tautomer enumeration with SMARTS-based plausibility scoring
• Dimorphite-DL protonation state enumeration over a pH window
• Henderson–Hasselbalch 5-layer microstate scoring
• PubChem experimental pKa evidence to guide state selection
• ETKDGv3 + MMFF 3D minimization for the top microstates
• Optional ML backends: pkasolver, propka, unipka CLI
• CLI (pkanet) and Python API
• Streamlit web UI (optional)

---

Installation

# Minimal (RDKit + requests only — heuristic mode)
pip install pkanet-cloud

# Recommended (adds Dimorphite-DL and pandas)
pip install "pkanet-cloud[recommended]"

# Full (adds propka and Streamlit web UI)
pip install "pkanet-cloud[all]"

RDKit note: rdkit is listed as a dependency but PyPI's rdkit wheel requires Python ≥ 3.9 on 64-bit platforms. If your environment uses a conda-managed RDKit, install without deps:

pip install pkanet-cloud --no-deps
pip install requests dimorphite-dl pandas   # then add these manually

---

CLI

After installation a pkanet command is available:

# Single SMILES
pkanet --smiles "CC(=O)OC1=CC=CC=C1C(=O)O" --ph 7.4

# Compound name  (PubChem auto-lookup)
pkanet --name "baicalein" --ph 7.4 --out-dir ./results

# SMILES file  (one SMILES [name] per line)
pkanet --smi-file ligands.smi --formats PDB MOL2 --top-n 3

# Ligand file  (.pdb / .mol2 / .sdf)
pkanet --file ligand.pdb --no-pubchem --keep-stereo

# Fast heuristic charge only  (no 3D, sub-millisecond)
pkanet --smiles "NCC(=O)O" --fast
pkanet --smi-file library.smi --fast --quiet   # TSV output

All CLI flags

Flag	Default	Description
--ph	7.4	Target pH
--out-dir	./pkanet_out	Output directory
--out-name	ligand	Base name for output files
--formats PDB MOL2 SDF	PDB	3D output format(s)
--ph-window	1.0	Dimorphite-DL pH window
--max-tautomers	8	Max tautomers to enumerate
--top-n	5	Top N microstates to rank
--top-k-3d	3	Write 3D for top-k microstates
--keep-stereo	off	Skip R/S stereoisomer enumeration
--no-pubchem	off	Disable PubChem pKa lookup
--fast	off	Heuristic charge only (no 3D)
--json-out FILE	—	Write full results as JSON
--quiet	off	TSV output, no banner

---

Python API

from pkanet import predict_charge, heuristic_net_charge, batch_predict_charges

# Auto mode: fast for unambiguous sites, full pipeline for borderline pKa
charge, mode = predict_charge("NCC(=O)O", ph=7.4)
print(charge, mode)   # 0  'full'

# Sub-millisecond heuristic estimate
charge = heuristic_net_charge("CC(=O)OC1=CC=CC=C1C(=O)O", ph=7.4)
print(charge)         # -1

# Batch prediction → pandas DataFrame
df = batch_predict_charges(
    ["NCC(=O)O", "CC(N)C(=O)O", "c1ccccc1O"],
    ph=7.4,
    mode="auto",
)
print(df[["name", "predicted_charge", "is_zwitterion"]])

Full pipeline

from pkanet.core import run_job

result = run_job(
    input_type    = "SMILES",
    smiles_text   = "CC(=O)OC1=CC=CC=C1C(=O)O",
    uploaded_bytes = None,
    uploaded_name  = None,
    target_pH     = 7.4,
    output_name   = "aspirin",
    out_dir       = "./output",
    output_formats = ["PDB"],
    use_pubchem   = True,
)

top = result["results"][0]
print(top["selected_microstate_smiles"])
print(top["formal_charge"])
print(top["minimized_pdb"])

---

pKa backends (auto-selected by priority)

Priority	Backend	Install
1	pkasolver (GNN)	pip install pkasolver
2	propka (semi-empirical)	pip install "pkanet-cloud[propka]"
3	unipka CLI	system install
4	heuristic SMARTS table	built-in, always available

---

Citation

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

Hengphasatporn K. et al. DFDD / Anyone Can Dock. JCIM (2026).
Ropp PJ et al. Dimorphite-DL. J Cheminform (2019).

---

License

MIT © Kowit Hengphasatporn — [e-mail]