Computational studies of anaplastic lymphoma kinase mutations reveal common mechanisms of oncogenic activation.

摘要: Kinases play important roles in diverse cellular processes, including signaling, differentiation, proliferation, and metabolism. They are frequently mutated cancer the targets of a large number specific inhibitors. Surveys genome atlases reveal that kinase domains, which consist 300 amino acids, can harbor numerous (150 to 200) single-point mutations across different patients same disease. This preponderance mutations-some activating, some silent-in known target protein make clinical decisions for enrolling drug trials challenging since relevance its sensitivity often depend on mutational status given patient. We show through computational studies using molecular dynamics (MD) as well enhanced sampling simulations experimentally determined activation be predicted effectively by identifying hydrogen bonding fingerprint loop αC-helix regions, despite fact occur throughout domain. In our study, we find predictive power MD is superior purely data-driven machine learning model involving biochemical features implemented, even though utilized far fewer (in fact, just one) an unsupervised setting. Moreover, results provide key insights into convergent mechanisms activation, primarily differential stabilization bond network engages residues active-like conformation >70% studied, regardless location mutation).