AI Blocks · All skills by job · Minds · Mini tools · How it works · From a source · Apps

Molecule property predictor

Give it molecules and measured results. Get a tested model that predicts a property, with fair train and test splits.

Molecule property predictor is a free AI skill for the job "Research a topic properly". Give the instructions to an AI assistant that supports the required files and tools, then review its output.

Free · No account · MIT licence · Made by K-Dense · Level: advanced · Thinking: high

Made by K-Dense. See the original on GitHub; we copied it as it is, with its licence.

What it does

A tested model that predicts a property of a molecule, such as solubility or toxicity, with fair splits so the score means something. The skill uses DeepChem. It first cleans your data: it parses molecule strings, records rejected rows and checks duplicate structures and repeated measurements. It trains a simple number-based baseline before any neural network. It splits by molecular scaffold or by time, so test molecules are truly new, and fits preprocessing on training data only. Settings are chosen on validation data and the final test is run once. The report shows per-task counts and warns that a good fit is not proof for new chemistry.

You give it

Molecules as SMILES strings
Measured values or labels
Units and assay conditions
The deployment question, such as new scaffolds (optional)

You get

A cleaning report with rejected rows
A baseline and a tuned model
A scaffold or time split
Test scores with counts per task
Limits on where to trust it

Use it when

You have lab results for a few thousand compounds and want to predict the rest.
Your model looks great and you worry about leakage between train and test.
You need a baseline before trying a graph neural network.
You want results in the original units, not scaled numbers.

How to use it

Add it to your AI once: download it and add the folder to Claude, ChatGPT or Codex as a skill.
Ask for the job in your own words. Your AI opens the skill when it fits.
It works best when your AI can run Python code.