liveBook · Manning

Kinodata

As explained in Build AI Drug Discovery Pipelines

Kinodata is a curated dataset of kinase bioactivity measurements assembled for computational kinase research.

Kinodata is a curated dataset of kinase bioactivity measurements for computational drug discovery. Its EGFR subset contains 7,287 curated activities used to train models that predict pIC50, a potency measure derived from IC50. section 8.1.1

A practical Kinodata workflow includes:

  • Use pIC50 rather than raw IC50 because the logarithmic scale reduces skewness; higher pIC50 indicates greater potency. section 8.1.1
  • Prefer the curated pIC50 values supplied by Kinodata; raw ChEMBL IC50 values may be stored as nanomolar standard_value measurements requiring conversion. section 8.1.1
  • Aggregate repeated measurements, such as by median pIC50, because compounds can recur across assays, laboratories, and publications. section 8.1.1

EGFR inhibitor discovery

Train an EGFR pIC50 prediction model with Kinodata, then screen the Enamine hinge binder library to select 100 promising and diverse molecules for downstream investigation. section 8.1.1

How Build AI Drug Discovery Pipelines covers it

How deep it goes
Defines curated measurements, pIC50 handling, duplicate aggregation, and assay noise, but does not detail dataset files or schema.
Examples
Uses 7,287 curated EGFR activities and screens the Enamine hinge binder library for 100 diverse candidates.
What it assumes
Assumes basic IC50 familiarity while defining pIC50, measurement units, duplicate assays, and curation choices.
What you can do after
Apply it via a predictive model to rank candidate EGFR inhibitors for downstream investigation.
How it connects
Connects kinase bioactivity with EGFR, ChEMBL, pIC50, IC50, and virtual screening.

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.