Enzyme-based synthesis and AI pocket analysis advance DNA-encoded libraries

Two new studies advance DNA-encoded library (DEL) technology for drug discovery. One team, from the University of Bern in collaboration with ETH Zurich and the Zurich University of Applied Sciences (ZHAW), developed an enzyme-based method that builds DNA-barcoded molecules under mild, water-based conditions without damaging the DNA tags. Another study introduced ErePOC, a pocket representation model based on contrastive learning that predicts which human proteins are suitable for DEL screening with 98% precision in downstream classification tasks.

DNA-encoded libraries (DELs) facilitate high-throughput screening of trillions of molecules against protein targets through split-pool synthesis and DNA tagging. In DEL technology, scientists synthesize an enormous number of different small molecules and attach a unique DNA 'barcode' to each one; these DNA-barcoded molecules can then be screened in parallel to determine which ones bind best to a disease-relevant protein. However, many of the chemical reactions traditionally used to build DELs are too harsh for the sensitive DNA tags and can damage them, and only a few DEL-derived compounds have advanced to clinical trials or reached the market.

To overcome the limitation of DNA-damaging reactions, the Bern-led team used two types of enzymes — CoA ligases and specially developed N-acyltransferases — to synthesize more than 120 diverse DNA-barcoded molecules under mild, water-based conditions without damaging the barcodes. Using protein engineering, the researchers tailored these enzymes to process molecules that already carry large, bulky DNA barcodes, and combined the two enzyme classes so that they carried out several reaction steps in sequence, functioning like a small production line. In a subsequent step, they linked the enzymatic reactions with classical chemical methods to assemble the DNA-encoded library directly on the DNA. The study, supported by the Swiss National Science Foundation (SNSF) as part of a Sinergia project, was published in Nature Catalysis.

“Enzymes are Nature’s catalysts: they accelerate reactions, work very precisely, and function in water under very mild conditions,” said the study’s lead author, a professor at the Department of Chemistry, Biochemistry and Pharmaceutical Sciences at the University of Bern. One of the two lead authors added that enzymes have long been known as versatile tools for making small molecules and are widely used in industry, yet until now they have hardly been used to build DNA-encoded libraries.

The separate study presented ErePOC, a pocket representation model based on contrastive learning with ESM-2 embeddings. ErePOC captures both structural and functional features of binding pockets, enabling identification of shared characteristics among DEL targets. With 98% precision in downstream classification tasks, ErePOC demonstrates high performance in pocket representation, and was applied to predict human proteins suitable for DEL screening, with enrichment uncovered across 18 functional categories. The work establishes a framework for enhancing DEL-based drug discovery through more effective target selection and pocket similarity analysis.

The enzyme-based approach could accelerate the search for new drugs and make the underlying chemistry more resource efficient. ErePOC provides a framework for enhancing DEL-based drug discovery through more effective target selection and pocket similarity analysis.