Check out our review on DNA-scaffolded catalysis! This link provides free full text access until the end of 2025: authors.elsevier.com/a/1m43W9CpcY.... Big thanks to co-authors @edwardpimentel.bsky.social , Ashley Ogorek, @ethan-hartman-125.bsky.social , and Caleb Cox

The enhanced bioorthogonality of the bCP ester enabled us to perform spatially-resolved RNA proximity labeling and RNA sequencing, including in human cells lines with high endogenous esterase activity.

9/n

The pCP / evolved BS2 pair performs well in multiple sub-compartments of mammalian cells.

8/n

We teamed up with @xuhuihuangchem.bsky.social to perform structural modeling and substrate docking to gain insights into the beneficial mutations.

7/n

Using DEEPMACh, we evolved BS2 esterase to increase activity toward pCP esters more than 230-fold.

6/n

To overcome this challenge, we developed a new platform, Directed Evolution of Enzymes via Masked Acid Chloride Probes ("DEEPMACh”). DEEPMACh combines yeast surface display with masked acylating probes, enabling rapid screening of >40 million enzyme mutants.

5/n

…BS2 esterase shows very low activity toward the pCP ester.

4/n

The methylcyclopropyl (mCP) ester protecting group together with BS2 esterase has been applied as a bioorthogonal system, but background unmasking of mCP occurs in mammalian cells. We found that the bulkier phenylCP group was much more bioorthogonal! However…

3/n

Context: combining bioorthogonal protecting groups with localized catalysts that unmask them is a powerful approach to modulate molecular activity. However, existing protecting groups are insufficiently bioorthogonal, or the catalysts that unmask them cannot be genetically targeted. 2/n

Excitingly, we observed correlation between ML predicted DNA nanoscaffold yields and experimental DNA-free reactions, including for kinetic time courses and for reactants not represented in the original DNA nanoscaffold library.

9/n

We teamed up with Matt Sigman and Beck Miller to use data science in library design and to combine our datasets with ML to generate predictive models. The model covers 18,000 reactant combinations under many reaction conditions.

8/n

Using our platform, one researcher performed >500,000 reactions in parallel with less than 3 days of bench time. By incorporating standards and calibration curves, we obtained precise DNA-scaffolded reaction yields using DNA sequencing.

7/n

Our approach draws inspiration from the pioneering work of David Liu on DNA-templated substrate coupling to screen for bond-forming reactions, www.nature.com/articles/nat... and the Vipergen Yoctoreactor which brings together 3 components for DNA-encoded small molecule synthesis.

6/n

Our platform works by 3 steps: 1) combinatorial self-assembly of a DNA nanoscaffold library, 2) Simultaneous scaffolded reactions and high-throughput selection of product-bearing nanoscaffolds 3) PCR amplification and next-generation DNA sequencing to identify successful reactions conditions.

5/n

We wondered whether we could use DNA scaffolding and next-generation DNA sequencing to dramatically increase the throughput, allowing >500,000 reactions to be performed in parallel, with simultaneous reaction analysis by sequencing.

4/n

Synergistic catalysis can unlock new reactivity, but it presents a multidimensional screening challenge. In the example scenario below, there are >500,000 combinations.

2/n