Randomized Subspace Learning for Proline Cis-Trans Isomerization Prediction

Proline Isomerization: From the Chemistry and Biology to Therapeutic Opportunities

Proline isomerization, the process of interconversion between the cis- and trans-forms of proline, is an important and unique post-translational modification that can affect protein folding and conformations, and ultimately regulate protein functions and biological pathways. Although impactful, the importance and prevalence of proline isomerization as a regulation mechanism in biological systems have not been fully understood or recognized. Aiming to fill gaps and bring new awareness, we attempt to provide a wholistic review on proline isomerization that firstly covers what proline isomerization is and the basic chemistry behind it. In this section, we vividly show that the cause of the unique ability of proline to adopt both cis- and trans-conformations in significant abundance is rooted from the steric hindrance of these two forms being similar, which is different from that in linear residues. We then discuss how proline isomerization was discovered historically followed by an introduction to all three types of proline isomerases and how proline isomerization plays a role in various cellular responses, such as cell cycle regulation, DNA damage repair, T-cell activation, and ion channel gating. We then explore various human diseases that have been linked to the dysregulation of proline isomerization. Finally, we wrap up with the current stage of various inhibitors developed to target proline isomerases as a strategy for therapeutic development.

Detecting Proline and Non-Proline Cis Isomers in Protein Structures from Sequences Using Deep Residual Ensemble Learning

It has been long established that cis conformations of amino acid residues play many biologically important roles despite their rare occurrence in protein structure. Because of this rarity, few methods have been developed for predicting cis isomers from protein sequences, most of which are based on outdated datasets and lack the means for independent testing. In this work, using a database of >10000 high-resolution protein structures, we update the statistics of cis isomers and develop a sequence-based prediction technique using an ensemble of residual convolutional and long short-term memory bidirectional recurrent neural networks that allow learning from the whole protein sequence. We show that ensembling eight neural network models yields maximum Matthews correlation coefficient values of approximately 0.35 for cis-Pro isomers and 0.1 for cis-nonPro residues. The method should be useful for prioritizing functionally important residues in cis isomers for experimental validations and improving the sampling of rare protein conformations for ab initio protein structure prediction.