Biotechnology · Intermediate

How to Prepare a Compound Collection Prior to Virtual Screening

Written by SignalsLab

Early edition: these lessons are being reviewed.

Content

Summary

Virtual screening is a computationally efficient alternative to high-throughput screening that evaluates large compound libraries to identify bioactive hits at a fraction of the cost. To ensure accuracy, chemical collections must be rigorously processed to remove unwanted structures, such as reactive species or aggregators, which can produce false positives in biological assays.

Intuition

Think of virtual screening as a massive digital filter that replaces the physical labor of high-throughput screening (HTS). While HTS has a median hit rate of around 0.3% and requires significant resources, virtual screening can evaluate millions of compounds quickly and often yields higher hit rates. However, this computational speed is only useful if the input data is clean. Just as a garbage-in-garbage-out principle applies, if you feed the system reactive species or aggregators, it will confidently identify them as 'hits' when they are actually just assay artifacts. Therefore, the core of the lesson is that preparation—filtering properties and removing promiscuous molecules—is the critical step that transforms a raw database into a reliable tool for drug discovery.

Derivation

  1. Virtual screening is defined as a computationally efficient alternative to HTS for evaluating large compound libraries.
  2. HTS typically has a median hit rate of around 0.3% and is resource-intensive.
  3. Virtual screening can evaluate millions of compounds at a fraction of the cost and frequently yields higher hit rates.
  4. Two major categories exist: structure-based (using 3D target structures) and ligand-based (using similarity searches).
  5. Unwanted structures like reactive species, aggregators, and frequent hitters cause false positives.
  6. Therefore, compound collections must be processed using property filters and cheminformatics tools before evaluation.
  7. ADMET profiling is often incorporated to assess drug-likeness.
  8. Big data from databases like PubChem and ChEMBL enhances predictive power.

Worked example

Problem

A research team runs a virtual screen on a library of 1 million compounds and identifies 50 'hits' that show strong activity in initial assays. However, follow-up experiments reveal these compounds are actually aggregators that interfere with the assay optics. What step was missed in the preparation workflow?

  1. Identify that the 'hits' were false positives caused by aggregators.
  2. Recognize that aggregators are explicitly listed as unwanted structures that cloud post-HTS analyses.
  3. Recall that proper preparation involves applying property filters and using online cheminformatics tools to remove such species.
  4. Conclude that the library was not rigorously processed to remove aggregators prior to screening.

Answer The team failed to rigorously process the chemical collection to remove aggregators and other unwanted structures before the virtual screen.

Common mistakes

  • Assuming that virtual screening results are always accurate without prior data curation; reactive species and aggregators can lead to false positives if not removed.
  • Confusing virtual screening with HTS; while both aim to find bioactive hits, virtual screening is computational and significantly cheaper, whereas HTS is experimental and resource-intensive.
  • Believing that virtual screening is only useful when no structural information is available; data fusion approaches can be used when both target structure and ligand information are present.

Animation

Animation coming soon

The animated version of this lesson is being prepared for the web. In the Signals AI app, lessons are drawn step by step.

Code

This is a text lesson: it has no animation code yet.

ccBYNC© 2026 SignalsLab. Lesson text shared as is under CC BY-NC 4.0.