Predicting Alternative Protein Conformations with AlphaFold2

Proteins are dynamic molecules capable of transitioning between open, closed, active, inactive, bound, and unbound conformations in accordance with the environment inside the cell. As a result, a given protein can have one conformation that is easily observable and another one which plays its role in signaling, transport, catalysis, or drug interactions. This paper investigates the methods of creating multiple protein conformations with the help of AlphaFold2 and AFsample2 software and suggests a two-step approach for selection of conformation based on the generated ensemble. Techniques like AF-Cluster, MSA Subsampling, SPEACH_AF, and AFsample2 alter the input of evolution to AlphaFold2 to uncover weaker structural signals. Apart from the output of AFsample2, an algorithm called CAPER (Context-Aware Prioritization of Ensemble Representatives) is proposed, which considers geometry, regulatory and molecular context, as well as completeness of the information. The approach presented should be understood not as a model of the populations or transition trajectories of conformations, but as a screening technique. The evaluation criteria for this technique should be able to differentiate two problems: whether AFsample2 can generate the alternative conformation at all, and whether CAPER ranks biologically relevant conformations highly once they have been generated.