Percepta: Comparing predicted and experimental values in one view

Description
When experimental data are available, users may want to compare Percepta predictions to measured values for calibration or model evaluation.

Solution

• Prepare a data table including:
  • Structures (SMILES, SDF).
  • Experimental values for the property of interest.
  • Identifiers (compound name or ID).
• Import this dataset into Percepta (where supported) and run predictions for the same property.
• View results:
  • Ensure both predicted and experimental columns are available.
• Export the combined results to CSV or Excel.
• In a spreadsheet or statistics tool:
  • Plot experimental vs predicted values.
  • Calculate correlation coefficients, RMSE, or bias as needed.
• Use this comparison to:
  • Assess model performance for your chemical space.
  • Identify outliers for further investigation.