SCMTPP

SCMTPP predicts thermophilic proteins from amino acid sequence using the scoring card method combined with g-gap dipeptide propensity scores to identify sequence determinants of thermostability.

Key Features:

• Novel methodology: Integrates the scoring card method (SCM) with estimated propensity scores derived from g-gap dipeptides to improve prediction accuracy and interpretability.
• High-quality dataset: Developed and evaluated on a dataset of 1853 thermophilic proteins and 3233 non-thermophilic proteins sourced from published literature.
• Performance metrics: Achieves cross-validation accuracy of 0.883 and is comparable to support vector machine (SVM) predictors (accuracy range: 0.906–0.910), outperforming other machine learning models by 2–17%.
• Independent-test performance: On an independent test set, attains accuracy of 0.865 and a Matthews correlation coefficient (MCC) of 0.731, outperforming ThermoPred.
• Interpretability: Provides insights into key physicochemical properties that contribute to protein thermostability.

Scientific Applications:

• Basic research: Facilitates analysis of sequence determinants and structural stability mechanisms of proteins at high temperatures.
• Industrial applications: Supports identification of thermophilic enzymes relevant to the food industry and other processes requiring high-temperature stability.
• Experimental guidance: Highlights physicochemical properties to guide protein engineering and experimental characterization of thermostability.

Methodology:

Applies the scoring card method (SCM) with estimated propensity scores from g-gap dipeptides; model training and evaluation used cross-validation and independent test sets on the 1853 TTP / 3233 non-TTP dataset, with benchmarking against SVM and other machine learning models and reporting of Matthews correlation coefficient (MCC).