Virtual Library of 14,300 Energetic Material Derivatives with ML Property Predictions

A virtual library of 14,300 derivatives based on the 8-nitro-2,6-dihydrotetrazolo[1,5-c]pyrimidin-5(3H)-one scaffold for energetic materials discovery. Predictive models for ten physicochemical properties were developed using Extreme Gradient Boosting (XGBoost) on density functional theory (DFT) data from 554 representative samples. The dataset was created by Jing Yang and last updated on 2026-05-26.

Use Cases

• Screen for high-energy-density candidates based on predicted detonation velocity and pressure.
• Identify insensitive material candidates based on predicted impact sensitivity (H50).
• Validate machine learning models for molecular property prediction in materials science.
• Explore structure-property relationships for the tetrazolo-pyrimidinone scaffold.

Strengths

• Contains a virtual library of 14,300 molecular derivatives.
• Predictive models were validated with DFT calculations, showing relative errors mostly below 2%.
• Models were trained on DFT-calculated data for 554 representative samples.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count for the primary data is unknown, which may limit suitability assessment.
• The dataset is small at 1.5 MB, which may indicate limited raw data.

Provenance

Source

figshare

Collection Method

Virtual library constructed using 23 energetic substituents; property predictions from ML models trained on DFT-calculated data.

Freshness

Last updated 2026-05-26 09:16:34; freshness should be verified.