MLysPRED

MLysPRED predicts multiple lysine posttranslational modifications (K-PTMs) in proteins, enabling identification of acetylation, crotonylation, methylation, and succinylation sites for functional analysis.

Key Features:

• Multi-Label Prediction Model: Predicts acetylation, crotonylation, methylation, and succinylation on lysine residues using a multi-label framework.
• Sequence Encoding Algorithms: Implements MLDBPB, MLPSDAAP, and MLPSTAAP encoding algorithms integrated with CHHAA (Contextual HMM-based Amino Acid), DR (Discrete Representation), and Kmer strategies to convert preprocessed lysine sequences into numerical features.
• Sampling Techniques: Addresses class imbalance via multidimensional normal distribution oversampling and a graph-based multi-view clustering under-sampling algorithm.
• Classification Algorithm: Uses a multi-label nearest neighbor algorithm for classification of lysine modification sites.
• Performance Metrics: Reported performance on independent datasets: Aiming 92.21%, Coverage 94.98%, Accuracy 89.63%, Absolute-True rate 81.46%, and Absolute-False rate 0.0682%.
• Comparative Evaluation: Includes comparative analyses against five existing predictors.

Scientific Applications:

• Basic Research: Enables study of protein function regulation by predicting multiple K-PTMs simultaneously.
• Drug Development: Facilitates identification of potential therapeutic targets through comprehensive lysine modification profiling.
• Proteomics Data Analysis: Supports analysis of complex K-PTM datasets in experimental proteomics.

Methodology:

Uses MLDBPB, MLPSDAAP, and MLPSTAAP encoding algorithms with CHHAA, DR, and Kmer strategies; applies multidimensional normal distribution oversampling and graph-based multi-view clustering under-sampling; and classifies with a multi-label nearest neighbor algorithm.