Run a first ZINC lookup by asking Ivy in chat: search by SMILES or ZINC ID, have Ivy filter by molecular weight and logP, then export the SMILES with their ZINC IDs.

Quick ZINC compound search with .bloom* platform

The ZINC database is a free resource of commercially available compounds for virtual screening, developed by the Irwin and Shoichet Laboratories at UCSF. ZINC-22, the latest major version, contains over 37 billion enumerated compounds in 2D and 750+ million purchasable compounds with 3D representations.

Citations

Tingle BI, Tang KG, Castanon M, Gutierrez JJ, Khurelbaatar M, Dandarchuluun C, Moroz YS, Irwin JJ. ZINC-22—A Free Multi-Billion-Scale Database of Tangible Compounds for Ligand Discovery. Journal of Chemical Information and Modeling. 2023;63(4):1166-1176. doi: 10.1021/acs.jcim.2c01253

Website: https://zinc.docking.org

Database highlights

Hundreds of millions of purchasable compounds in ready-to-dock 3D formats
37+ billion compounds searchable in 2D (make-on-demand libraries)
Includes catalogs from Enamine REAL, WuXi GalaXi, and Mcule Ultimate
Free access via CartBlanche22 interface, Globus, AWS, and Oracle OCI clouds

Step-by-step compound library creation

There is no ZINC form or search page in .bloom*. You run a ZINC lookup by asking Ivy in chat, and it runs inside that turn — no job to track.

Ask in chat

Find purchasable ZINC compounds that exactly match CC(=O)Oc1ccccc1C(=O)O
Find close ZINC analogs of gefitinib and list their vendor codes
Look up ZINC000000000001 and check vendor availability
Which ZINC compounds match catalog code EN300-19380?

What the lookup can do

Search by SMILES: an exact match, or close analogs by graph edit distance (1-4 is the useful range; wider distances return looser matches). Up to 1,000 results, 100 by default.
Look up ZINC IDs: up to 100 at once, returning SMILES, vendor codes and catalog names.
Check vendor catalog codes: up to 100 at once, returning the matching ZINC IDs and SMILES.
Fields returned: ZINC ID, SMILES, tranche, molecular weight and logP by default; ask for reactivity, purchasability or features if you need them.

The lookup does not filter by property range, run a Tanimoto similarity search, or match a SMARTS pattern. Ivy does those steps in the chat's code workspace with RDKit, on the compounds the lookup returned or on a bulk pull from ZINC's CartBlanche22 service (see Property Filters and Downloads & Batching).

Review and export

Results come back in the chat turn as a compound list with ZINC IDs, SMILES and vendor codes
Ask Ivy to save the list as a CSV, or to convert it to 3D SDF for docking (3D Prep & SDF Exports)
Ask Ivy to send the compounds on to another tool, such as DiffDock or ADMET-AI

Common ZINC search scenarios

Each scenario below combines a ZINC lookup with property and structural-alert filtering that Ivy applies in the code workspace. Describe the scenario in chat and Ivy runs both parts.

Drug-like compound library

Set MW: 150-500 Da, logP: -2 to 5, TPSA: <140
Choose "in-stock" subset for immediate availability
Enable PAINS and reactivity filters
Export 1000-5000 compounds for virtual screening

Fragment-based screening

Set MW: 100-300 Da for fragment-sized molecules
Adjust logP and solubility properties
Use substructure search for specific scaffolds
Focus on high-quality, well-characterized fragments

Natural product-like diversity

Use higher MW ranges (300-600 Da)
Include more complex ring systems
Apply natural product-inspired filters
Sample diverse chemical space for novelty

ADMET-optimized compounds

Set properties favorable for drug development
Filter for good solubility and permeability
Exclude compounds with known liabilities
Generate focused libraries for hit-to-lead

Quality control and validation

Search result validation

Compound quality:

Check SMILES validity (use structure preview)
Verify molecular properties match expectations
Review for unusual or reactive functional groups
Confirm compounds are in appropriate subset

Data completeness:

Ensure all requested properties are present
Check for missing values in critical fields
Validate ZINC IDs for uniqueness
Review timestamp for data freshness

Search optimization

Performance considerations:

Start with smaller result sets for query testing
Use property filters early to reduce search scope
Consider search complexity vs. result quality trade-off
Save successful search parameters for reuse

Result quality assessment:

Use diversity metrics to ensure chemical space coverage
Check for duplicate or very similar compounds
Validate against known active compounds if applicable
Consider follow-up filtering with other tools

Integration with .bloom* workflows

Downstream processing

Virtual screening preparation:

Send ZINC compounds directly to DiffDock
Use for molecular docking campaigns
Prepare libraries for virtual screening workflows

Property prediction:

Submit to ADMET-AI for comprehensive profiling
Use DeepSol for solubility prediction
Generate multi-property datasets

Compound management:

Import into .bloom*'s compound database
Create focused screening libraries
Track compound provenance and metadata

Research applications

Hit identification:

Generate diverse screening libraries
Create target-focused compound sets
Design novelty-oriented discovery libraries

Lead optimization:

Find analogs of known active compounds
Generate focused series for SAR studies
Identify backup compounds for development

Method development:

Test new screening methodologies
Benchmark virtual screening tools
Validate computational chemistry approaches

Troubleshooting common issues

No results returned

Broaden property ranges or remove restrictive filters
Check subset availability (try different stock categories)
Simplify substructure queries if too specific
Verify SMILES/SMARTS syntax

Poor compound quality

Enable PAINS and reactivity filters
Use more restrictive property ranges
Apply substructure filters for specific chemotypes
Consider alternative search strategies

Performance issues

Reduce requested compound count
Add more property filters to limit scope
Use simpler search queries
Consider batch processing for large requests

Data interpretation

Understand subset meanings (in-stock vs. make-on-demand)
Consider pricing and availability implications
Review property calculation methods
Validate against known compounds when possible

Best practices for ZINC usage

Search strategy

Combine property and substructure filters for optimal results
Save successful search parameters for reproducible queries
Document search criteria for research reproducibility
Consider cost vs. diversity trade-offs

Data management

Use consistent naming conventions for exported files
Include metadata about search parameters in file headers
Version control important compound sets
Maintain provenance information for research tracking

Workflow integration

Plan compound processing pipelines before large searches
Consider downstream tool requirements when selecting compounds
Batch process large datasets to manage computational resources
Automate repetitive search and filtering operations

Chat Usage

Example Prompts

"Find close ZINC analogs of gefitinib"
"Find close ZINC analogs of gefitinib, then keep MW 200-400 and LogP < 3"
"Look up ZINC000000000001 and check vendor availability"

Workflow Chaining

Feeds into: ADMET screening, DiffDock (docking), AiZynthFinder (synthesis), Compound Passport
Accepts from: PubChem (SMILES), REINVENT/GenMol (generated molecules for analog search)

What Ivy can look up in ZINC22

By SMILES: exact match or close analogs by edit distance
By ZINC ID: SMILES, vendor codes and catalog names
By vendor catalog code: the matching ZINC IDs and SMILES (prices come from a separate vendor price search)

ZINC Quickstart