liveBook · Manning

hERG inhibition

As explained in Build AI Drug Discovery Pipelines

hERG blockage is inhibition of the human ether-à-go-go-related gene potassium channel by a chemical compound.

hERG inhibition is blockage of the human ether-à-go-go-related gene potassium channel by a chemical compound. In ligand-based screening, it becomes a binary prediction problem: determine whether a compound blocks the channel from molecular features and historical experimental labels. section 3.2.1 section 3.3.1

A computational hERG-inhibition workflow uses:

  • SMILES or other molecular representations as inputs, including molecular descriptors and structural fingerprints. section 3.3.1
  • Experimental activity labels, with 1 indicating channel blockage and 0 indicating no blockage. section 3.2.1
  • A supervised model that learns feature-to-label weights and is later evaluated on previously unseen compounds. section 3.3.1

Threshold-based labels

The dataset defines a compound as a hERG blocker when IC50 is below 40 μM, equivalent to pIC50 above 4.4. Because higher pIC50 means greater potency, a value of 9.0 represents stronger inhibition than a value of 6.0. section 3.2.1

Listing 3.1 Loading hERG Blockers Dataset
from pathlib import Path
import pandas as pd
import urllib.request
 
def load_herg_blockers_data():
  herg_blockers_path = Path("data/ch03/hERG_blockers.xlsx")
 
  return pd.read_excel(
      herg_blockers_path,
      usecols="A:F",
      header=None,
      skiprows=[0,1],
      names=[
          "SMILES", "Name", "pIC50",
          "Class", "Scaffold Split", "Random Split"
      ],
    ).head(-68)    #A

Load the 587-compound dataset into Pandas, retaining SMILES, compound names, pIC50, class labels, and random or scaffold split assignments. section 3.2.1

How Build AI Drug Discovery Pipelines covers it

How deep it goes
Defines IC50, pIC50, labels, features, training, and inference, but stops short of a complete model implementation.
Examples
Uses a 587-compound dataset, pIC50 thresholds, SMILES strings, and labeled example compounds such as DEMETHYLASTEMIZOLE.
What it assumes
Teaches IC50, pIC50, SMILES, labels, features, training, and inference before applying them to hERG prediction.
What you can do after
Evaluate/understand how molecular data and experimental labels support binary hERG-blockage prediction.
How it connects
Connects hERG inhibition with toxicity screening, molecular descriptors, structural fingerprints, supervised learning, and scaffold splitting.

This explainer was generated by Manning's knowledge graph from the text of Build AI Drug Discovery Pipelines. The section links point to the parts of the book it draws on.