Enhanced Prediction of Lysine Propionylation Sites using Bi-peptide Evolutionary Features Resolving Data Imbalance

Lysine propionylation is an extensive post- translational modification in both eukaryotes and prokaryotes which brings a significant regulatory role. However, for various biological processes and metabolic activities, it is also responsible. Nevertheless, the limitation and effectiveness of lysine propionylation in photosynthetic organisms remain obscure. If any computational method can thoroughly identify the propionylation sites with greater accuracy then it will be more beneficial as well as save the money and cost. In this work, we have introduced a novel feature extraction method to predict the propionylation sites using the bi-peptide based evolutionary feature concept. Afterward, we have applied the Support Vector Machine (SVM) classifier that gave Matthew's correlation coefficient of 0.72, Sensitivity of 73.05 %, Specificity of 97.05 %, Accuracy of 85.05 %, and F1-score of 0.82 when using 10-fold cross-validation. Our proposed model outperforms the previously developed tool, PropPred, in all cases.