Recently, a big leap forward in predicting protein structures came with the development of AlphaFold2, and similar deep learning methods. However, AlphaFold2 has been shown to have difficulties with flexible proteins. Professor Maya Topf’s team (Integrative Virology) in collaboration with Kay Grünewald (Structural cell biology of viruses) and Kostas Thalassinos (at University College London), proposed a method (XLMS-tools) for combining AlphaFold2 with crosslinking mass spectrometry (XL-MS data to identify the experimentally-relevant conformers from an ensemble of protein structures. Published in Molecular & Cellular Proteomics, XLMS-tools employs AlphaFold2 to generate potential structures at different conformations, and XL-MS data to identify the most relevant one under different conditions.
"Many proteins exist as multiple structures, depending on what is attached to the protein, or other external cues,”
said Kish Adoni, one of the collaborators from the Thalassinos lab. “XL-MS is a versatile and sensitive technique for elucidating how these structures change under different circumstances." Specifically, XL-MS provides distance constraints and surface exposure information, which, when combined, effectively constrain protein structures.
"By itself, each individual piece of information from XL-MS does not uniquely identify the structure of a particular protein," explained Karen Manalastas-Cantos, the main developer of the method.
"It’s like solving a mystery where, by process of elimination, we can use each datapoint to iteratively narrow down the field to the most likely suspects – or in our case, the most likely protein structures.”
The workflow was shown to effectively identify the correct protein structure using simulated and experimental data. For example, in the case of glutamine-binding periplasmic protein (QBP) the monolinks helped to determine the open conformation. According to Maya Topf:
“This is particularly important, as monolink information has thus far been neglected in favour of crosslinks in structure modelling.”
The source code for this workflow is available at https://gitlab.com/topf-lab/xlms-tools
Comments