RF-MaloSite and DL-Malosite: Methods based on random forest and deep learning to identify malonylation sites

Area of Science:

• Biochemistry
• Computational Biology
• Proteomics

Background:

• Lysine malonylation is an emerging post-translational modification regulating biological activities.
• Malonylation is implicated in diseases such as cardiovascular disease and cancer.
• Current methods for identifying malonylation sites are often time-consuming and technically challenging.

Purpose of the Study:

• To develop novel computational methods for predicting lysine malonylation sites.
• To provide efficient and accurate alternatives to experimental proteomics analysis for malonylation site identification.

Main Methods:

• Development of two machine learning-based prediction tools: RF-MaloSite (Random Forest) and DL-MaloSite (Deep Learning).
• RF-MaloSite utilizes biochemical, physiochemical, and sequence-based features.
• DL-MaloSite uses the primary amino acid sequence as input.

Main Results:

• Both RF-MaloSite and DL-MaloSite demonstrated high performance across various metrics, including accuracy and sensitivity.
• DL-MaloSite achieved Matthew's Correlation Coefficient (MCC) scores of 0.51 (cross-validation) and 0.49 (independent test set).
• RF-MaloSite achieved MCC scores of 0.42 (cross-validation) and 0.40 (independent test set), with both methods showing competitive or superior efficiency compared to existing tools.

Conclusions:

• RF-MaloSite and DL-MaloSite are effective computational tools for predicting lysine malonylation sites.
• These methods can accelerate research into the biological roles of malonylation and its involvement in disease.
• The findings may offer insights into crosstalk between malonylation and other lysine modifications.