Tgprediction

The system integrates a large curated dataset of over 22,000 polymer glass transition temperatures and applies multi‑level feature engineering to extract 34 physicochemical descriptors from SMILES representations. A gradient‑boosted regression model, trained on both high‑quality baseline data and a transfer‑learning bridge set of hydrogen‑bonding polymers, delivers end‑to‑end predictions directly from DNA or RNA sequences. Users can interactively input sequences or run batch predictions via a command‑line interface, receiving Tg estimates along with confidence scores. The tool also provides SHAP‑based feature importance visualizations and supports exporting results in JSON for downstream workflows. Designed for polymer scientists and materials researchers, it accelerates the discovery of high‑performance polymers and aids in selecting materials for temperature‑critical applications.