DNA-encoded fragment growing

When growing fragments into leads, the typical route is making and purifying one compound at a time. In recent years parallel synthesis and screening of crude reaction mixtures has been catching on. However, no physical screening approach can match the throughput of DNA-encoded libraries (DEL). In a new (open access) Chem. Sci. paper, Michael Waring and collaborators at Newcastle University, University of Oxford, and Genentech use DEL to grow fragments.

 

The researchers started by creating a “poised DNA-encoded library.” We’ve previously discussed the concept of poised libraries, which are designed for rapid follow-up chemistry. Typically, the library members are fragments that contain a handle to facilitate growing. Here, the concept is reversed, with the DEL itself poised to react with a pre-chosen fragment. In this first test case, the DEL consisted of just 42 members made by coupling 7 amino acids to 6 aryl halide-containing acids, which could then be used for Suzuki-Miyaura couplings.

 

Bromodomains such as the BD1 domain of BRD4 bind a plethora of published fragments, and the researchers chose the 7-atom 3,5-dimethylisoxazole, one of the first fragments published. A boronic acid version of this was coupled to the DEL library and screened against BD1. One DNA sequence in particular occurred 11-times more frequently than any of the 41 others. The corresponding molecule was synthesized without being attached to DNA and found to have a dissociation constant of 51 nM as assessed by NMR. Three control molecules which used different amino acid or aryl-halide building blocks had affinities considerably lower, the best being 2.5 µM.

 

 

A crystal structure of compound 22 bound to BD1 showed several important contacts that explain why the molecule was selected in the DEL screen. Moreover, a more lipophilic version of the molecule showed some cell-based activity.

 

The “NUDEL” (for Newcastle University DEL) is an interesting approach to rapidly explore regions of chemical space around a fragment hit. By including every possible combination of building blocks in the library it is possible to find synergistic combinations, such as those found in compound 22; molecules derived from alanine but a different aryl halide or pyrazole but a different amino acid were not selected over background levels.

 

Of course, as the researchers acknowledge, 42 compounds is very modest for a DEL. Considerable care was taken to ensure each library member was properly synthesized to facilitate proper analysis (for example, that the selection was not based on differences in concentrations of different library members). This level of care would be more difficult with a million compound library. Also, finding high affinity binders to BRD4 is a rather low bar, particularly when starting with a known fragment. Nonetheless NUDELs look like they could prove quite useful, and I look forward to seeing applications to more novel targets. Perhaps they could even be combined with the DEL-based fragment finding approach we highlighted last year. I predict growing bonds between fragments and DEL.