GERD Drug Compound Model Predictions with Residuals

Predicted and observed property values for Gastroesophageal Reflux Disease drug compounds enable performance evaluation of QSAR/QSPR models. The dataset includes molecule identifiers, target properties, and residuals for predictions from various algorithms and descriptor sets. Author Mythili V published the dataset on figshare in April 2026.

Use Cases

• Benchmark model performance by comparing `Predicted Value` against `Actual Value` across different `Model Name / Algorithm` entries.
• Analyze systematic prediction errors by examining the `Residual (Error)` column grouped by `Target Property` or `Descriptor Set`.
• Identify compounds where models fail by filtering for high-magnitude `Residual (Error)` for specific `Molecule Identifier` entries.
• Compare the suitability of different `Descriptor Set` inputs for predicting specific `Target Property` values.
• Validate new QSAR models by using the `Actual Value` column as a reference benchmark for specific GERD drug compounds.

Strengths

• Includes direct error metrics via the calculated `Residual (Error)` column for each prediction.
• Facilitates model comparison by listing the `Model Name / Algorithm` and `Descriptor Set` used for each prediction.

Limitations

• The total number of compounds, prediction instances, and specific property ranges are unknown.
• Potential for model or data source bias is unknown, as the origin of the `Actual Value` (experimental or reference) is not detailed.

Provenance

Source

figshare, author Mythili V.

Collection Method

Generated from machine learning and statistical models developed for QSAR/QSPR analysis.

Freshness

Last updated in April 2026.