Protein microstructuralism after AlphaFold

Abstract The success of Google DeepMind’s AlphaFold in predicting protein structures has profound implications for the debate between microstructuralism and functionalism in the philosophy of biochemistry. This paper addresses whether protein kinds are distinguished by their internal microstructure, their functional role, or both: a question complicated by moonlighting proteins that perform multiple functions and adopt various shapes. While Tobin (2010) argues that moonlighting proteins challenge microstructuralism due to their lack of a single, stable tertiary structure, this paper defends Goodwin’s (2011) counterargument. Goodwin contends that a protein’s stable primary amino acid sequence defines its conformation space - the finite set of physically stable structural states accessible to it. Recent advancements in artificial intelligence lend significant weight to this view. In successfully predicting over 200 million protein structures, AlphaFold demonstrates that structural outcomes are rule-based and follow from the primary sequence. I argue that the potential for AI to map a protein’s conformation space from its sequence alone suggests that functional versatility is encoded within fundamental primary structure. This supports a form of ontological monism: despite their complexity, proteins are defined by one essential property - their primary structure. Finally, I contend that this microstructuralist framework remains robust even when faced with the erratic behaviour of pathological amyloids, which are better classified as distinct proteinaceous deposits rather than protein kinds.