Publication Details
SoluProt: Prediction of Protein Solubility
Marušiak Martin, Ing.
Martínek Tomáš, doc. Ing., Ph.D. (DCSY)
Zendulka Jaroslav, doc. Ing., CSc. (UIFS)
Bednář David (FIT)
Damborský Jiří, prof. Mgr., Dr. (UMEL)
protein, solubility, prediction, machine-learning
Protein solubility poses a major bottleneck in production of many therapeutic and industrially attractive proteins. Experimental solubilization attempts are plagued by relatively low success rates and often lead to the loss of biological activity. Therefore, any advance in computational prediction of protein solubility may reduce the cost of experimental studies significantly. Here, we propose a novel software tool SoluProt for prediction of solubility from protein sequence based on machine learning and TargetTrack database. SoluProt achieved the best accuracy 58.2% and AUC 0.61 of all available tools at an independent balanced test set derived from NESG database. While the absolute prediction performance is rather low, SoluProt can still help to reduce costs of experimental studies significantly by efficient prioritization of protein sequences. The main SoluProt contribution lies in improved preprocessing of noisy training data and sensible selection of sequence features included in the prediction model.
@inproceedings{BUT155085,
author="Jiří {Hon} and Martin {Marušiak} and Tomáš {Martínek} and Jaroslav {Zendulka} and David {Bednář} and Jiří {Damborský}",
title="SoluProt: Prediction of Protein Solubility",
booktitle="DAZ & WIKT 2018 Proceedings",
year="2018",
pages="261--265",
publisher="Brno University of Technology",
address="Brno",
isbn="978-80-214-5679-2",
url="https://www.fit.vut.cz/research/publication/11808/"
}