Reaction Optimisation in Drug Discovery

Traditionally, medicinal chemistry projects involve the preparation of as many products as possible, especially at the beginning stages. Sometimes the yields and purities obtained in specific, critical reactions are low, and there is ample room for improvement.

Medicinal chemistry, however, is about new compounds, not perfect chemistry.

In the last year we approached our clients with a new idea: In some cases, optimisation can be carried out parallel to the main work, and while one group of chemists prepares new compounds, a second group can improve the chemistry. This second group’s results can then be applied to the synthesis in progress. The goal is to add an external resource which can offer new value in a short period of time.

One of our clients approached us with a request to optimize the preparation of an heterocyclic scaffold used as a starting point for a drug discovery route. The product was obtained through the SNAr reaction of a chloroheterocycle with a high-value amine. Although the reaction worked, it was slow, conversion was low and a dimeric impurity was present in significant quantities. These problems translated into low yields (34% after 3 days) and difficult purifications.

Time was of paramount importance. A fast bibliographic search revealed that similar reactions had been described using metal couplings, but not over our scaffold. Therefore, to avoid longer development times, metals were discarded.

Although the application of microwave conditions would be an obvious improvement, this alternative was also discarded to avoid scale-up issues. We concentrated our efforts instead on finding a new combination of solvent and base, first designing a fast-screening process to evaluate nine different solvents and two bases, and using the % of starting material and product obtained by LC-MS as response factors.

Not all 18 possible combinations were considered, since clues within the bibliography led us to expect poor results from some combinations. Rather, an initial set of six reactions was carried out in parallel in a Radley’s carousel.

Initially, all reactions were carried out at 86 °C with 1.5 eq of base. The reactions were monitored at different times, and soon it was clear that more base was needed. Two additional eq of the corresponding base therefore were added to each reaction.

Some of the conclusions drawn were:

• EtOH and PrOH with DIPEA (entries 1 and 2) gave similar results, with a 50% conversion after 52 h. No dimeric product was detected.
• CH₃CN with DIPEA (entry 3) gave a mediocre result, with a 27% conversion after 52 h, but traces of dimeric product were detected.
• BuOH (entry 4) presented the least favorable results, with increasing quantities of the dimeric compound being detected in each sample.
• DCE with K₂CO₃ (entry 5) was a no-go, with no conversion.
• CH₃CN with K₂CO₃ (entry 6) gave acceptable results, although the conversion was lower than that using DIPEA. No dimeric product was detected.

EtOH and PrOH results were promising. CH₃CN also had potential, with a good conversion at a shorter reaction time. The next batch of experiments was set up to include other polar solvents, with 3.5 eq of base from the beginning. Additionally, since conversion was similar at 44h and 52 h, a maximum time was established at 44 h.

The results from the second set of experiments were much more promising:

• Entries 7, 8 and 14 revealed that other polar solvents did not improve the CH₃CN result.
• Entries 9 and 10 revealed that a full quantity of base from the beginning resulted in no improvement with alcohols.
• Entry 11 revealed similar behaviour for CH₃CN with DIPEA; namely no improvement.
• Entry 12 revealed that for CH₃CN with K₂CO₃, a full quantity of base from the beginning makes a difference.
• Entry 13 revealed that pyridine, which is both solvent and the base, yielded no product.

Clearly entry 12, 3.5 eq of K₂CO₃ in CH₃CN for 42 h at 86 °C provided the best results. However, the reaction time could be shortened. A sample of the reaction monitored by LC-MS at 24 h showed a 16:83 ratio—not much less favorable than the final 13:87. We therefore determined that a 1-2% increase in conversion was not worth the 24 additional hours needed to obtain it. The reaction was finally reproduced on a larger scale with similar results: 12:87 ratio at 26 h.

We provided our client with these conditions, detailed in a full report of the experiment, which included each set of results. Whilst the reaction was not complete after 26 h, no dimeric product was detected and purification was much easier than under previous conditions.