Exploring secondary structural transitions in microtubule severing motors with machine learning

Katanin is a microtubule severing nanomachine that becomes fully functional in a hexameric assembly driven by binding to co-factors; however, it has been observed to populate lower order oligomeric states. Literature findings suggest that the hexameric state of katanin is prone to disassembly into lower order oligomers and furthermore that this disassembly may be accompanied by a secondary structure change in the protomers. To gain insight into the molecular factors driving these two processes, we carried out long all-atom molecular dynamics simulations on various order oligomeric states of katanin in the presence or absence of binding partners. To characterize changes to the secondary structure of the protomers, we developed a centroid-based clustering method applied directly to the Ramachandran plot to analyze structural changes occurring in simulations. We evaluated the performance of our method through the application of established clustering methods to our simulations and of our methodology to existing simulations of intrinsically disordered proteins, which are known to populate varied secondary structures. We will discuss the results of our analysis, which highlight significant conformational transitions and their link to underlying allosteric networks in oligomers.