Discussion

Within the pharmaceutical industry the vast majority of the early phases of drug discovery are performed using solubilised samples, predominantly in di-methyl sulfoxide (DMSO). Whilst DMSO is a very good solvent for “drug like” molecules a small percentage have limited, or no, solubility and these samples, if not detected, can lead to blockages in liquid handlers and misleading results in subsequent assays.
Across our global sites local operating procedures had been developed to deal with precipitating samples, including heating, sonication and the addition of water, and whilst some of these procedures solved the immediate problem of dissolving the sample they generated other issues around degradation or even that the sample re-precipitated shortly afterwards.
Despite the fact that a low percentage of compounds precipitate in DMSO we had identified several thousand over the many years of compound management operations and these structures were used by our Computational Chemistry department to look for chemical features that occur predominantly in insoluble compounds. Using this knowledge they were able to build a predictive model that could score the risk of a given structure having DMSO solubility issues, which when applied to real samples allowed us to identify >95% of the precipitates by only visually checking around 10% of the solubilised samples.
Using our global expertise and the knowledge gained one solution was devised, incorporating the use of the prediction tool and standardising the process globally. The meant we could improve handling efficiency, reduce risk and improve quality for our customers.