Building a Knowledgebase
The key to building a chiral knowledgebase lies in automation. Since most organizations are limited by instrument availability and manpower, it is critical to set up a system to automatically collect the knowledge during day-to-day operations. Fortunately, the experiment-intensive nature of the task faced by the purification lab provides a wealth of experimental data to form the basis of the knowledgebase. The challenge is efficient capture of the information.
The minimum information necessary to form a unit of knowledge should be obvious—the structure, the method, and the effectiveness of that method (often measured in terms of enantiomeric peak resolution). Typically, chromatographic information, including chemical structures, chromatograms, and peak resolutions, are scattered across CDSs (chromatographic data systems), registries, electronic notebooks, and paper or PDF reports. This information must be automatically compiled to form the knowledgebase.
Scriptable automation components designed to interface with various analytical and structure-based systems can import chromatograms, and where possible, instrument parameters from native CDSs such as ChemStation. Queries on ELNs using metadata, such as sample names, can retrieve chemical structures.
Configured to collect, compile, and evaluate data as it is generated, the system can monitor methods rated by effectiveness. With little to no additional manual effort from users, information is stored in an analytical database supporting chromatographic, structural, and metadata content.
Fast user review of the chromatographic ratings is streamlined through the use of a customizable table view (Figure 2). Visually, the chromatographer can confirm the accuracy of the rating, revising as necessary, and flag the most viable method for scale-up. Thus in a few seconds, a complete knowledgebase can be compiled for a given project compound.
To effectively predict chiral selectivities for a given method, it is necessary to accumulate a complete variety of selectivities on the compounds of interest. The temptation may be to retain only the effective method used during scale-up, providing some insight into the effectiveness of a given method for a given compound. However, to maximize the value of the knowledgebase, at least semi-quantitative values are required—good, fair, and bad. These can form the basis of a system designed to indicate the probability of success of a method for a new structure.