Impact Sensitivity Predictions for 404 Energetic Nitro Compounds

A quantitative structure–property relationship (QSPR) dataset for predicting the impact sensitivity (logH50) of 404 nitro compounds. The dataset was created by Yali Xu and published on figshare in May 2026. It contains results from multilayer perceptron and 1D-CNN models trained on Dragon molecular descriptors.

Use Cases

• Training QSPR models for impact sensitivity prediction based on Dragon descriptors.
• Benchmarking deep learning model performance (MLP vs. CNN) for chemical property prediction.
• Analyzing the influence of electronic distribution and molecular bulk on sensitivity based on the mechanistic interpretation.
• Designing safer nitro compounds by leveraging the provided physicochemical insights.

Strengths

• Models achieved high test set performance (R² of 0.853 for MLP, 0.843 for CNN).
• Dataset includes 404 nitro compounds, providing a basis for modeling structurally diverse molecules.
• Both models satisfy all QSPR acceptance criteria, as stated in the description.
• Mechanistic interpretation provides insight into governing factors like polarity and hydrogen bonding.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.
• The dataset is small in scale at 81.7 KB, indicating limited raw data volume.

Provenance

Source

figshare, author Yali Xu.

Collection Method

Quantitative structure–property relationship (QSPR) modeling using Dragon descriptors and deep learning.

Freshness

Last updated 2026-05-27 19:40:03; freshness should be verified.