RF-MaloSite and DL-MaloSite

RF-MaloSite and DL-MaloSite predict protein lysine malonylation sites to identify post-translational modification positions implicated in biological processes and disease and to complement tandem mass spectrometry analyses.

Key Features:

• Algorithmic foundation: RF-MaloSite uses a random forest algorithm while DL-MaloSite employs deep learning techniques.
• Feature representation: RF-MaloSite utilizes a comprehensive set of biochemical, physiochemical, and sequence-based features; DL-MaloSite requires only the primary amino acid sequence as input.
• Performance metrics: Evaluation includes accuracy, sensitivity, and Matthew's Correlation Coefficient (MCC), with RF-MaloSite MCC scores of 0.42 (10-fold cross-validation) and 0.40 (independent test set) and DL-MaloSite MCC scores of 0.51 (10-fold cross-validation) and 0.49 (independent test set).
• Evaluation protocol: Both methods were assessed using 10-fold cross-validation and an independent test set.
• Comparative efficiency: The methods perform on par with or better than existing malonylation site prediction approaches.

Scientific Applications:

• Regulatory-role analysis: Predicting malonylation sites to investigate the regulatory roles of lysine malonylation in cellular processes.
• Cross-talk analysis: Enabling analysis of interactions between malonylation and other lysine modifications such as acetylation, glutarylation, and succinylation.
• Disease mechanism elucidation: Supporting studies that link malonylation patterns to molecular mechanisms underlying diseases including cardiovascular disease and cancer.

Methodology:

Computational methods include a random forest classifier using biochemical, physiochemical, and sequence-based features for RF-MaloSite, deep learning models using primary amino acid sequence for DL-MaloSite, and evaluation by 10-fold cross-validation and an independent test set using accuracy, sensitivity, and MCC.