1074 LCGC VOLUME 18 NUMBER 10 OCTOBER 2000 www.chromatographyonline.com

Chromatography Data Systems in Drug Discovery

The author discusses the roles of mass spectrometry and chromatography data systems within drug discovery analytical laboratories. He also presents ways of using laboratory information management systems to integrate the chemical data.

n this article I’ll examine the use of • liquid chromatography–mass spectrome- chromatography data systems in drug try (LC–MS), which is used to determine discovery. Of particular interest is ana- compound identity based on the predic- I lyzing the results of the chemical soups tion from the anticipated compound to be produced by combinatorial chemistry labo- synthesized ratories. • nuclear magnetic resonance spectroscopy, The problem is that chromatographers which serves as a backup for compound working in combinatorial chemistry can be identification and is used mainly at the in a no-win situation: the combinatorial start of the library synthesis. chemistry increases throughput, and the I’ll concentrate on HPLC and MS analy- result is that the number of samples to be sis and on the data systems used to manage assayed also increases. How can this work the data produced. If analysts in small labo- flow be managed? What can chromatogra- ratories synthesize more than 100,000 com- phers do to reduce the waiting times? One pounds per year, they generally will need to idea, presented by Sage and co-workers (1), interpret one chromatogram and one spec- is to use parallel analysis with an eight- trum per compound. Still printing on paper? channel multiplexed electrospray interface That’s a lot of trees. to increase sample throughput. I discussed this topic in overview in a pre- vious publication (2), but I want to concen- General Directions in trate on the data analysis aspects in more Drug Discovery detail. I’ll be looking specifically at the chro- The general trends in drug discovery can be matography data systems used within labo- summed up as more, faster, and higher: ratories analyzing combinatorial chemistry more compounds synthesized faster than products. before and with higher purity. This pace will Analytical data flows in drug discovery: enable the synthesized compounds to be Figure 1 shows the data flows that come used in high-throughput screening quicker through analytical laboratories. The library than before, and the samples should last is planned, starting chemicals are sourced, longer than they did previously. and the synthesis is planned at the level of To achieve this process, the chemical iden- the individual 96-well plate. At this point tity and purity must be known. Analytical the informatics software can pass informa- laboratories and chromatographers provide tion about the expected compound to be this information. synthesized in each plate well. This informa- tion is used as the basis for calculating the Analysis in Drug Discovery expected mass ion for the compound in each The main analytical techniques in drug dis- individual well; the data will be compared covery and their goals are with the resulting mass spectrum. R.D. McDowall • high performance liquid chromatography The samples in the plate wells are ana- McDowall Consulting, 73 Murray (HPLC), which is used to determine lyzed first by MS. This step analyzes the Avenue, Bromley, Kent BR1 3DJ, purity using area normalization (percent- products in the solution and confirms that United Kingdom age of the peak area of total area counts) the identity on the premise of the mass ion 1075 LCGC VOLUME 18 NUMBER 10 OCTOBER 2000 www.chromatographyonline.com is equivalent to the expected molecular in light of the speed of analysis and turn- Figure 2 illustrates how the data systems weight. Some adduct ions will be produced around time. would communicate via LIMS. The applica- under certain ionization conditions will After completing the MS and HPLC tion controlling the synthesis of the combi- require manual interpretation by analysts, analyses, workers must collate the results. If natorial chemistry library with information but these situations are relatively few in the data systems can communicate effec- about the anticipated identities of com- number compared with the numbers of tively, the transfer and collation of results pounds to be synthesized would send data to compounds synthesized. For MS analysis, should be relatively easy. As discussed above, the LIMS. The plate identities and well posi- the turnaround time can be limited by the they usually don’t, so a third application tions would be identified and downloaded speed of the autosampler feeding the instru- must be used to collate the results and pro- into the MS system. After analysis and inter- ment, which will have a wash cycle that duce a report of the day’s analyses, as well as pretation by the MS data system and ana- could be in the range of 1 min or so. The an overall library when the whole analysis lysts, the plate details about the wells with MS analysis itself can be over within half of has been completed. positive hits would be uploaded into the that time. Potential solutions with data handling: LIMS. Wells that were not worth analyzing Data handling with MS software is rela- Many of the problems with data handling in would be identified and downloaded to the tively straightforward. The software can pro- this area have arisen because of the rapid HPLC data system. The LC analysis would duce a report to identify compounds that development of combinatorial chemistry skip those wells and assay only those with have been correctly synthesized and those and the associated analysis. The introduc- the appropriate compound present, thus sav- wells that have failed identification. tion of commercial software products has ing time. This approach would save time The next stage is the HPLC gradient been slow in this area, so many solutions are over the common analyze-everything analysis. The goal of this step is a short gra- developed in-house. approach outlined above and would allow dient that can separate the compounds used The first problem with data handling is the chromatography resources to be used for in the synthesis from the anticipated new that the MS and the chromatography data more added-value analyses. chemical entities. This process will require systems must communicate and transfer Is a commercial LIMS worth installing? some preliminary method development data and information between all systems. The answer is yes. The informatics problem work to ensure the column and mobile- Few systems can perform these tasks, and in laboratories is relatively simple compared phase conditions are capable of achieving they are major requirements to prevent ana- with a quality control laboratory in pharma- this goal. The gradients are relatively steep lytical laboratories from becoming larger ceutical manufacturing. Each combinatorial and recycled to the starting point relatively bottlenecks than they are already. This com- chemistry compound yields two main pieces quickly. The complication of this step is that munication will need a better third-party of information (ignoring repeat analyses): it requires 5–6 min. solution than a spreadsheet. Perhaps one molecular weight (identity) and purity. The To save time, analytical chemists ideally solution can be the use of a commercial lab- reports from the system must be based should assay only those samples correctly oratory information management system around this information with a link to the synthesized; that is, the samples that demon- (LIMS). raw data files produced by both instruments. strated the correct mass ion from the MS analysis. There is little point analyzing sam- ples that you already know do not contain your required compound. However, trans- ferring the plate information from the MS data system to the data system running the HPLC can be problematic because these sys- tems usually are manufactured by different vendors and may not talk to each other eas- ily. Due to this communication problem, some laboratories may need to assay the whole plate for purity regardless of the MS results, because it is the easiest way to oper- ate. This process obviously wastes time and effort, so there is a need for the data systems to communicate with each other. The results of a gradient run can be ana- lyzed using an area percentage normalization method. The total peak area between defined points is used to calculate the per- centage purity. This determination is fine if the lambda maximum of the compound is the same as the wavelength used in the detector, but this outcome usually is not the case. Therefore, the calculation usually will have some inaccuracy, but it can be ignored Figure 1: Process flow for analysis of combinatorial chemistry libraries. www.chromatographyonline.com OCTOBER 2000 LCGC VOLUME 18 NUMBER 10 1076 Specifications of the compounds are rela- The major difference comes from the upward by at least an order of magnitude tively simple: identity and purity compared sheer numbers of compounds that are syn- over the next 2–3 years, if expansion con- with the quality control laboratory that may thesized and must be managed by the tinues at the current rate. have 10 and 20 analyses with in-house and LIMS. The system must be able to cope manufacturing specifications. with the current numbers and be scalable Conclusion Reducing and, ideally, eliminating the use of paper in drug discovery analytical labora- tories can be approached by careful design and integration of the data systems. Usually MS and chromatography data systems are incompatible, but the data from each sys- tem can be integrated effectively in a LIMS. The LIMS itself can integrate with other systems that plan the combinatorial chem- istry synthesis and plate designs to ensure rapid turnaround of results.

References (1) A.B. Sage, D. Little, and K. Giles, LCGC, Cur- rent Trends and Developments in Drug Discovery 18(5S), S20–S29 (2000). (2) R.D. McDowall, LCGC, Current Trends and Developments in Drug Discovery 18(5S), S8–S13 (2000). Ⅲ

Figure 2: The integration of separate MS and chromatography data systems with a LIMS.

C