AlphaFold3 and protein modifications: obvious strengths and significant challenges

Proteins are fundamental macromolecules in cells, serving a diverse range of functions. Understanding how cells orchestrate transitions between states is crucial for understanding organismal development. Proteins undergo various post-translational modifications (PTMs), such as phosphorylation, acetylation and ubiquitination, which dynamically alter their structure, function, folding, and interactions. Dysregulation of these modifications can disrupt cellular homeostasis, leading to diseases such as cancer and Alzheimer's, and affecting protein proteostasis. Despite their significance, our understanding of the structure-function relationship of proteins and PTMs is very limited due to limited experimental data on protein structures, the diversity of PTMs, and the dynamic nature of proteins. Deep-learning methods such as AlphaFold and RoseTTAFold have revolutionized protein structure prediction and design, yet their current capabilities are limited to specific systems like covalent modifications and mutations. The latest releases, RoseTTAFold-All-Atom and AlphaFold 3, demonstrate a significant increase in accuracy in predicting protein complexes, including nucleic acids, small molecules, and post-translational modifications. In this work, we highlight the strengths, weaknesses, and major challenges associated with state-of-the-art protein structure prediction methods in relation to protein modifications, using phosphorylation as a case study.