Stability of biomolecules, especially of proteins, is of great interest and significance. Protein stability has been the major target for protein engineering, mainly to increase the stability, but sometimes also to destabilize proteins . Effects on stability are among the most common consequences for disease-related variations , thus this phenomenon is of interest for variation interpretation to explain the effects of harmful variants.

We performed a thorough check of the details in ProTherm and corrected numerous problems. In the end we had less than 50% of the original number of variants left. Out of these, 77% came from ProTherm, the rest are either corrected or new variants. With this high quality dataset we trained a novel machine learning predictor for amino acid substitution effects on stability and established a new baseline for variant stability prediction method performance.

Single variation prediction Multiple variations prediction About Disclaimer

Protein sequences file:
in FASTA format, e.g.


Protein variations file:
The information in each line include: variation_number, variation, protein_name, temperature, PH. e.g.

1 G71S gi115114 72.60 4.60
2 R36A gi115114 39.00 7.00
3 P37A gi115114 39.00 7.00
4 D38A gi115114 39.00 7.00

Email: (The prediction results will be sent to the supplied mail-box soon.)