DNA-encoded libraries (DELs) have been making a big splash in drug discovery recently. Although the original idea dates back to 1992, DELs seem now to be surpassing traditional high-throughput screening approaches in terms of cost, scale, and speed and promise to make it easy for small labs to screen billions of compounds against targets of interest. Similar to other in vitro display technologies, like phage display or mRNA display which our lab and the Pearce lab have used extensively, each compound in a DEL is tagged with a unique DNA barcode, so hit deconvolution can be achieved by routine DNA sequencing. Still, a heavy reliance on just a handful of chemistries, in particular peptide couplings and reductive aminations, have largely dragged DELs back into the chemical space of phage and mRNA display libraries. It seems surprising, given the wealth of classical organic reactions and newer, C-H modalities, that DEL chemistry should be dominated by these two. Numerous groups have begun to work on translating reactions such as the Suzuki coupling, Wittig olefination, olefin metathesis, or Diels-Alder into DNA-compatible forms, but substrate scope can still be a limiting factor and few of these reactions have been successfully applied in real selection campaigns.
Thus, we were very interested to read the new paper from the Baran and Blackmond labs at Scripps, which aims to develop a roadmap for translating known organic reactions into DEL-compatible chemistry. This work is presumably driven in part by an exciting recent partnership between Scripps and Pfizer for work on this front. In the paper, the authors first identified a set of general parameters that should be met for successful DNA-compatible chemistry and then, using the Giese reaction as a case study, they launch into an exhaustive six step kinetic and mechanistic analysis, involving screening of over 100 reaction conditions, that eventually lands them at the doorstep of DEL-compatible chemistry. Interestingly, the most significant discovery, in terms of optimizing the reaction is that concentration and available surface area of the stoichiometric Zn activator can be tuned to allow the use of dilute concentrations of other coupling partners and facilitate chemistry in the presence of DNA. Overall, this optimized Giese reaction provides a robust protocol to forge valuable C(sp3)-C(sp3) bonds on DNA templates, which the authors demonstrate with a broad substrate scope analysis.
There is little doubt that this chemistry is cool and could lead to some interesting new DELs and one obvious next step is to generate a DEL library with this chemistry and use it in a selection. But this paper really got our lab talking about what is a “DEL-compatible” reaction? For example, we noticed significant variation in the yields for this chemistry, ranging from 44-93% on model DEL substrates, and wondered what’s the minimum acceptable yield for DEL chemistry? A 40% yield may be too low. Consider if the average yield of each step in a 3-step synthesis is 40%, this would result in a library where only 6.4% is really viable, thus dropping the real diversity and making hit deconvolution very challenging. A minimum yield of 80% may be more realistic since >50% of the library will be present as intended. We were obviously not the first to worry about this: the authors cite no less than five different methods for determining yields of on-DEL chemistry from the lit. Significantly, Paegel and co-workers considered the impact that reaction conditions might have on the informational integrity of the DNA: a reaction may proceed smoothly in a given set of conditions and the DNA may still be intact, but is the DNA amplifiable? Paegel’s went beyond the conventional organic standard for reaction yield and measured an “amplifiable DNA remaining percentage,” which in many cases differs from the reaction yield measured by LCMS, indicating that informational integrity is a very real concern. We also wondered what would be the impact of mixtures of degenerate stereoisomers in a DEL library? Because the reaction, as presented in the paper, isn’t stereoselective, certain substrates give yield to epimers. If these epimers were present in a library, coded by the same tag, it’s unclear whether this would be beneficial because it would contribute to diversity or detrimental, since one isomer might drown out the selection of the other, especially if substrate control winds up leading to uneven mixtures of diastereomers.
These potential issues highlight the need for intense scrutiny when developing DNA-compatible reactions to yield a toolbox that won’t compromise the library in the process of making it. Therefore, it may be worth a moderate revision of the list of general reaction parameters to increase the minimum acceptable yield and include assessments of stereoselectivity and DNA-integrity or even measuring the impact of the latter two when adapting chemistries to DEL conditions. That being said, this paper represents a major step forward in translating classic organic reactions to DEL-compatible chemistry. Indeed, it strongly recommends designing algorithms to aid discovery of DNA-compatible reactions by chemical robots: adapting this paper’sapproach computationally, to search powerful resources like Scifinder or Reaxys and systematically comb through reaction conditions that fit the above parameters may be a quick way to identify new candidate reactions for translation into the DEL medium. Additionally, our lab loves using enzymes to do difficult and interesting chemistry, and DEL-compatible reaction conditions seem well suited for enzymes; we imagine that enzymes will become useful reagents for generating interesting DELs in the future.