| Summary: | 碩士 === 國立陽明大學 === 生物醫學資訊研究所 === 100 === Matrix metalloproteinases (MMPs) belong to a metzincin superfamily which is characterized by the HEXXHXXGXXH zinc-binding motif and Met-turn formed by conserved methionine. It has been shown that MMPs are capable to degrade most components of the extracellular matrix (ECM) as well as growth factors. During cancer progression, MMPs are required for tumor cells migration via the degradation of the ECM. In addition, in recent studies, MMPs are also been reported not only cleavage ECM but also shed membrane proteins into extracellular environment. With diverse substrates on MMPs-mediated cleavage, MMPs play a vital role in numerous biology pathways and physiological function. As a consequence, identifying numerous MMPs substrates seems to be a feasible approach to assess MPs-mediated biology function. In this study, a new tool ClevagePred is developed to predict MMPs substrates cleavage sites to find novel substrates. The approach is by constructing a computational prediction model based on support vector machine (SVM) and artificial neural network (ANN) with a 5-fold cross-validation. The prediction performance, with different cutting window sizes, was evaluated by accuracy, sensitivity, specificity, and Matthews correlation coefficient (MCC) measures. Based on results of this study, the predictive performance for SVM model is optimized at the widow size is 10 (P6 –P’4 ), with accuracy of 75.83%(SE:1.00%) and MMC of 0.517(SE:0.020). Te predictive performance for ANN model is optimized at the widow size 7 (P4 –P’3 ), with accuracy of 74.44%(SE:3.71%) and MMC of 0.492(SE:0.070). It is anticipated that ClevagePred can be served as a powerful tool for predicting substrate candidates.
|
|---|
