Markov mean properties for cell death-related protein classification

Area of Science:

• Computational biology
• Biochemistry
• Bioinformatics

Background:

• Cell death (CD) is crucial in physiological and pathological processes.
• Identifying CD molecular targets requires efficient theoretical methods.
• Current methods lack speed and accuracy in predicting CD-related proteins.

Purpose of the Study:

• To develop the first classification model for predicting cell death (CD)-related proteins.
• To utilize Markov Mean Properties and protein 3D structural information for prediction.
• To identify novel CD-related proteins with potential therapeutic applications.

Main Methods:

• Calculated protein descriptors using MInD-Prot with topological information, physicochemical properties, and 3D regions.
• Employed Machine Learning algorithms (Weka) for classification model development.
• Utilized feature subset selection to optimize the predictive model, identifying the K* algorithm as most accurate.

Main Results:

• Achieved a highly accurate classification model (K*) with an Area Under the Receiver Operating Characteristic Curve (AUROC) of 0.992 and a true positive rate of 88.2% on the validation set.
• Successfully predicted CD-related functions for several previously uncharacterized proteins in Homo sapiens and bacteria.
• Identified specific proteins (e.g., 3DRX, 4DWF, 1IUR, 1J7D, 1UTU, 3EEC, 2G3V, 4G5A, 1YLK, 1XSV) with predicted roles in various cellular processes and disease pathways.

Conclusions:

• Demonstrated the feasibility of predicting cell death (CD)-related proteins using molecular information derived from protein 3D structures.
• The developed model offers a rapid and accurate approach for identifying novel CD molecular targets.
• This work opens avenues for discovering new therapeutic strategies targeting cell death pathways.