Apoptosis proteins have a central role in the development and homeostasis of an organism. These proteins are very important for understanding the mechanism of programmed cell death. Based on the idea of coarse-grained description and grouping in physics, a new feature extraction method with grouped weight for protein sequence is presented, and applied to apoptosis protein subcellular localization prediction associated with support vector machine. For the same training dataset and the same predictive algorithm, the overall prediction accuracy of our method in Jackknife test is 13.2% and 15.3% higher than the accuracy based on the amino acid composition and instability index. Especially for the else class apoptosis proteins, the increment of prediction accuracy is 41.7 and 33.3 percentile, respectively. The experiment results show that the new feature extraction method is efficient to extract the structure information implicated in protein sequence and the method has reached a satisfied performance despite its simplicity. The overall prediction accuracy of EBGW_SVM model on dataset ZD98 reach 92.9% in Jackknife test, which is 8.2-20.4 percentile higher than other existing models. For a new dataset ZW225, the overall prediction accuracy of EBGW_SVM achieves 83.1%. Those implied that EBGW_SVM model is a simple but efficient prediction model for apoptosis protein subcellular location prediction.

Protein p K a prediction is essential for the investigation of the pH-associated relationship between protein structure and function. In this work, we introduce a deep learning-based protein p K a predictor DeepKa, which is trained and validated with the p K a values derived from continuous constant-pH molecular dynamics (CpHMD) simulations of 279 soluble proteins. Here, the CpHMD implemented in the Amber molecular dynamics package has been employed ( Huang Y. Huang Y. 29949356 J. Chem. Inf. Model. 2018 58 1372 1383 ). Notably, to avoid discontinuities at the boundary, grid charges are proposed to represent protein electrostatics. We show that the prediction accuracy by DeepKa is close to that by CpHMD benchmarking simulations, validating DeepKa as an efficient protein p K a predictor. In addition, the training and validation sets created in this study can be applied to the development of machine learning-based protein p K a predictors in the future. Finally, the grid charge representation is general and applicable to other topics, such as the protein–ligand binding affinity prediction.

No abstract

No abstract