liveBook · Manning

High-throughput screening

As explained in Build AI Drug Discovery Pipelines

High-throughput screening is the automated testing of many compounds in parallel for activity against a biological or chemical assay.

High-throughput screening (HTS) tests many compounds in parallel against a biological assay, using robotics and miniaturization to automate thousands of small experiments. It helps identify hit compounds that significantly inhibit or otherwise affect a target, before more detailed follow-up. Screening commonly begins with one concentration for rapid activity triage; promising compounds can then undergo testing across multiple concentrations. section 1.4.1 section 7.1.1

A practical HTS workflow typically includes:

  • Select a large compound library and an assay that measures the target activity of interest. section 1.4.1 section 2.0
  • Measure each compound at a relatively high single concentration, flagging compounds with significant inhibition as potential hits. section 7.1.1
  • When resources permit, test multiple concentrations and fit a dose-response curve to estimate IC50, the concentration producing half-maximal inhibition. section 7.1.1
  • Use the resulting measurements to train machine-learning models that expand experimental evidence into virtual screening. section 2.0

COVID Moonshot screening campaign

The COVID Moonshot campaign synthesized more than 2,000 designs and used HTS assays to measure Mpro inhibition. For 2,062 synthesized compounds, IC50 values came from dose-response experiments spanning 11 concentrations from 100 to 0.0017 μM. A computational workflow represented the molecules with 2,048-bit Morgan fingerprints and marked compounds below 5 μM IC50 as active for additional analysis. section 7.1 section 7.1.1

Table 2.1 Comparing high-throughput screening and virtual screening.
High Throughput Screening
Virtual Screening
Test 104 to 106 compounds
Test 108 to 1012 compounds (ultra-large libraries)
Can only screen compounds that have been synthesized
Can screen compounds that have not yet been synthesized
Generates beneficial, real data
Limited data generation
Expensive
Cheap
Requires an assay for experimental measurement of interest
Requires a computational model or rule set; ML-based approaches additionally need sufficient training data
Screening a large library takes days to weeks, including assay development and logistics
Can screen millions of compounds in minutes to hours once a model or workflow is in place

Compare HTS with virtual screening across library size, cost, assay requirements, data generation, and turnaround time. section 2.0

How Build AI Drug Discovery Pipelines covers it

How deep it goes
Explains library testing, single-shot triage, dose-response curves, IC50 fitting, and machine-learning reuse of assay data.
Examples
Uses kinase and cytochrome P450 inhibition, plus COVID Moonshot Mpro assays and IC50 measurements.
What it assumes
Assumes basic target and compound terminology while defining hits, inhibitors, assays, dose-response curves, and IC50.
What you can do after
Evaluate assay activity, flag follow-up hits, estimate IC50 values, and prepare HTS data for computational screening.
How it connects
Connects experimental screening with virtual screening, drug discovery, enzyme inhibition, dose-response curves, IC50, and machine learning.

Related concepts

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.