pka_predictor

Predict pKa values for small molecules from SMILES inputs.

52|11|Updated Mar 10, 2026

Search Tags:#prediction #chemistry #molecule #backends #pka #uni-pka

Ready to run

npx skills add https://github.com/InternScience/ChemClaw --skill pka-predictor

💬 Prompt for ChatGPT / Claude Chat▼

Please help me install this Agent Skill.
Skill: pka_predictor
Source: https://github.com/InternScience/ChemClaw/tree/main/skills/pka-predictor
Command: npx skills add https://github.com/InternScience/ChemClaw --skill pka-predictor

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill requires numpy, pandas, scipy, rdkit, torch, huggingface_hub, joblib, and includes scripts (resource) and assets (resource) components.

What problem does it solve?

pKa Predictor provides a unified entry point to estimate the acid dissociation constants (pKa) of small molecules, enabling researchers to quickly assess ionization properties for drug design and cheminformatics tasks.

Core Features & Use Cases

Dual backend support: Use a custom heuristic backend or Uni-pKa single-weight backend for flexible deployment.
Batch and single input: Accepts individual SMILES or batches; outputs structured results suitable for downstream pipelines.
Use Case: Compare pKa profiles across multiple drug-like molecules to prioritize lead candidates and understand protonation behavior under physiological pH.

Quick Start

Install dependencies and run the predictor with a SMILES input to obtain pKa estimates.

Frequently Asked Questions about pka_predictor

High-intent search queries and answers about installing and using this skill.

FAQPage Schema

How do I predict pKa values for small molecules using SMILES inputs?▼

The predictor estimates acid dissociation constants (pKa) by accepting SMILES strings as input, supporting both single-molecule and batch processing to yield structured results for downstream cheminformatics pipelines.

What is the difference between the custom heuristic and Uni-pKa backends for pKa prediction?▼

The custom heuristic backend offers flexible deployment for pKa estimation, while the Uni-pKa backend uses a single-weight model architecture requiring downloaded weights via HuggingFace CLI for small-molecule ionization prediction.

Do I need PyTorch and RDKit installed to run pKa predictions?▼

Yes, running pKa predictions requires Python dependencies including NumPy, PandAs, SciPy, RDKit, and PyTorch installed, plus downloading Uni-pKa weights via the HuggingFace CLI for the Uni-pKa backend.

Can I compare pKa profiles across multiple drug-like molecules in batch?▼

Yes, the predictor accepts batch SMILES inputs to compare pKa profiles across multiple drug-like molecules, enabling you to prioritize lead candidates and analyze protonation behavior under physiological pH.

Why use a pKa predictor for drug design and cheminformatics tasks?▼

Using a pKa predictor provides a unified entry point to estimate acid dissociation constants, enabling researchers to quickly assess small-molecule ionization properties required for drug design and cheminformatics tasks.

What are the limitations of using a heuristic backend for pKa estimation?▼

The custom heuristic backend provides flexible pKa estimation but may lack the advanced predictive modeling of the Uni-pKa single-weight backend, which leverages PyTorch and downloaded HuggingFace model weights.