---
title: ChEMBL similarity search
description: Find structurally similar compounds in ChEMBL using Tanimoto similarity.
tags:
  - chembl
  - similarity
---

Similarity search finds compounds in ChEMBL that are structurally similar to a query molecule, ranked by Tanimoto coefficient — useful for expanding SAR around a known active or hopping to new scaffolds. The threshold you ask for most affects the result: high values (≥85%) return few, near-identical hits, while lower values (50–70%) surface more diverse scaffold-hopping candidates.

## ChEMBL Similarity Search

Find compounds in ChEMBL that are structurally similar to a query molecule, ranked by Tanimoto similarity.

## What you can ask for

| Filter | Default | Limits |
|--------|---------|--------|
| Query molecule | Required | One SMILES string per search (not a comma-separated list), up to 1000 characters |
| Similarity threshold | 70% | Minimum Tanimoto similarity, 40-100% |
| Number of results | 20 | Up to 100 per search |

## Tanimoto Coefficient

The Tanimoto coefficient measures structural similarity between molecules based on molecular fingerprints. The threshold is a percentage from 40 to 100:

| Threshold | Interpretation |
|-----------|---------------|
| 100% | Identical fingerprints |
| ≥85% | Very high similarity (likely same scaffold) |
| 70–85% | High similarity (good for SAR expansion) |
| 50–70% | Moderate similarity (scaffold hopping candidates) |
| <50% | Low similarity (different chemical series) — thresholds below the 40% floor are rejected |

## Scope

Similarity search finds molecules that are globally similar by Tanimoto fingerprint comparison. There is no separate ChEMBL substructure-search tool on the platform — for a specific chemical motif, run this similarity search at a lower threshold (50-60%) to surface a broader candidate set, then filter the results by hand.

## Chat Usage

### Example Prompts

- "Find compounds similar to aspirin in ChEMBL with Tanimoto > 0.8"
- "Search ChEMBL for analogs of CC(=O)Oc1ccccc1C(=O)O above 0.7 similarity"

### Workflow Chaining

- **Feeds into**: ADMET screening, DiffDock, Compound Passport
- **Accepts from**: PubChem (compound SMILES), REINVENT (generated molecules)