Anindya, Atsarina Larasati2024-09-172024-09-172024-09-17978-91-8069-921-1 (Print)978-91-8069-922-8 (PDF)https://hdl.handle.net/2077/83197Proteins are typically viewed as a sequence of amino acids, where the sequence dictates folding into specific shapes to carry out distinct biological tasks by interacting with specific partners as a consequence of their complementary interfaces. This tidy sequence-to-function model interpreted through protein 3D structure, however, does not tell the whole story. Some proteins are moonlighters, juggling multiple, seemingly unrelated jobs, possibly aided by their ability to change shapes, move around the cell, and team up with different partners. The delicate dance of forming and breaking protein complexes hinges on two main factors: stability and specificity, both of which are fine-tuned by the balance between the solvent environment and weak atomic interactions. While we often think of protein interactions as straightforward, step-by-step processes, there is a growing recognition that a more dynamic approach, like using energy landscapes, might better capture the complexity—especially when it comes to these moonlighting proteins. The focus of this work is to take a part in unraveling protein moonlighting behavior by using Bayesian approach to analyze interaction kinetics and developing survivin binding prediction models to enhance our understanding of specificity, using survivin as the model protein due to its remarkable functional versatility. Bayesian progress curve analysis applied on survivin ligand-target interaction and survivin dimerization experiments using biolayer interferometry and microscale thermophoresis (MST) reduces selection bias and reveals interesting kinetic processes which otherwise can be potentially ignored. The binding prediction model shows that chemical group compositions contained in amino acids is a major decision factor for survivin interaction partner recognition, even without taking sequence or structure into account. This work also describes how the prediction model can be applied on the human proteome. The roles of specific amino acid residues on survivin dimerization are further explored with MST and small-angle X-ray scattering experiments.engbiochemistrymachine learningprotein interactionsbayesian progress curveprotein interaction predictionsurvivinText