Thyroid Peroxidase Inhibitor Classification Using SMILES Embeddings

A dataset used to develop quantitative structure-activity relationship (QSAR) models for classifying thyroid peroxidase (TPO) inhibitors. The research compares traditional molecular descriptors with SMILES embeddings generated via a pretrained Bidirectional Encoder Representations from Transformers (BERT) architecture. The dataset, 69.9 KB in size, was created by Geven Piir and last updated on June 2, 2026.

Use Cases

• Classifying TPO inhibitors based on SMILES string representations.
• Comparing model performance between traditional molecular descriptors and transfer-learned embeddings.
• Evaluating the applicability domain of QSAR models for agrochemical screening.

Strengths

• Dataset size is 69.9 KB, indicating a focused and manageable scope for model development.
• Research demonstrates a direct comparison between two modeling approaches (traditional descriptors vs. SMILES embeddings).
• External validation involved comparison with TPO inhibition data from EFSA and EU-NETVAL laboratories.

Limitations

• Row count is unknown, which may limit suitability assessment.
• Column-level documentation is absent; field semantics must be inferred after download.
• The description notes conflicting experimental data from some sources, suggesting uncertainty for some compounds.

Provenance

Source

figshare

Collection Method

Likely contains experimental and computational data on chemical compounds and their TPO inhibition activity.

Freshness

Last updated 2026-06-02 11:12:24; freshness should be verified.