Propionylation Characterization Ensembling Classifiers on Varied Features

A recently discovered post-translational modification (PTM) called propionylation where propionyl group is added in the lysine amino acid’s side chains causing diversity and evolution. This alteration occurs in both eukaryotic and prokaryotic cellular environments under a variety of circumstances. It plays a significant part in metabolic functions, the control of biomolecules and several biological processes with consequences for a number of unhealthy disorders. One approach is to use mass spectrometry methods that distinguish propionylation PTMs in a laboratory setting. However, this procedure is characterized by a slow pace and significant cost inefficiency. The development of computational models has the potential to replace traditional laboratory techniques, increasing its availability to users of all skill levels worldwide. We developed a simplified efficient technique that combines positional and evolutionary data from classical machine learning with VGG-16 and RNN architectures applied to semantic embedding features taken from natural language processing. We were able to obtain accuracy, sensitivity along with specificity and MCC of 96.53 %, 97.27 %, 94.56 % and 0.53 % respectively which fared better than the most recent state-of-the-art models built.