NCATS Logo NCATS Chemical Genomics Center
Free software and code for public use
NCGC CurveFit
Large scale dose response curve fitting and curve classification software

Download: Application. (requires Java WebStart)
Download: Sample Data File.
Download: Source Code.
Software Features
curvefit mainview
  • Automated curve fitting and classificatoin software
  • Algorithm recognizes bell shaped curves, implements standard Hill equation, extensible for other models; distinguishes activation vs inhibition
  • Stand alone tool designed explicitly for public use and for source code reference
  • Analyzes 10k curves with good performance, capacity to handle >100k curves with memory usage on user machine being the limit
  • Provides activity ranking of complete and incomplete curves
  • Fast chemical similarity and substructure searching (including smarts support) enabled using path-based fingerprints
  • Ability to export results, curve images
  • Web deployed software, keeps users current with latest features
curvefit r-group table
  • R-Group table generation using substructure querying
  • Software implements in-memory embedded database to allow for advanced querying/ filtering across all data in software
  • Java based code provided as a reference and the potential for integration & automation in workflow
  • Source code highlights:
    • Curve fitting algorithm that handles more curves than any commercial package (scalability outpaces screening capacity; millions of samples possible)
    • Curve classification method
    • R-group labeling algorithm
    • Curve rendering
    • Advanced and resuable query/retievable interface
    • Reusable and extensible Java library for handling databases & high performance molecular spreadsheets
    • JChem path-based fingerprint genration and use
Getting started:
Step 1. Launch application using link above, "Add Data" to load sample data.
Step 2. "Perform Fitting", allow software to finish batch processing.
Step 3. Sort by Activity Rank (sorted by inhibitors, activators and inactives). Class -1.1 is a full inhibitor, -1.2 partial inhibitor, -2.1 and -2.2 are partial curves, -3 are single point actives, 4 are inactives. Class 1.1 are full activators, etc. Class 4s sometimes have logAC50s, but these should be treated as inactives.
Step 4. Ctrl-click or shift-click to select multiple samples.
Step 5. Select Tools > Generate Fingerprint to create path-based fp that enables chemical similarity searching.
Step 6. Try Query function to filter results. For example, try searching on Curve Class = -1.1 to see best inhibitors.
Step 7. Query by smiles to perform similarity or substructure searching.
Step 8. To builid an R-group table, provide a core scaffold (you can optionally provide "R1", "R2" positions to be specific in scaffold definition)
Contact: Please contact Yuhong Wang for issues related to this software.