Elixabete Rezabal

For readability, detailed background is deferred to the Supporting Information . In the main text, we focus on two representative methods, SD and Adam, which offer a clear contrast between a classical gradient-based scheme and a modern adaptive optimizer. Further discussion of CG, AdaGrad, and RMSProp is also provided in the Supporting Information .

1 Introduction The utilization of advanced materials as platforms to support functional proteins, such as enzymes, offers multiple benefits, including enhanced enzyme stability against denaturation, increased resistance to inhibitory compounds, and simplified separation and reuse of biocatalysts [1-7].

Encouraging developments demonstrate that few transition metal and organometallic catalysts can operate in a bioorthogonal fashion and promote non-natural chemistry in living systems by minimizing undesired side reactions with cellular components. These catalytic processes have potential for applications in medicinal chemistry and chemical biology. Nevertheless, the stringent conditions of the cell environment severely limit the number of accessible metal catalysts and exogenous reactions.