Towards Automated Evaluation Of Result Accuracy For LC/MS/UV/ELSD/CLND Substance Screening – Supporting Library Management And Medicinal Chemistry
ASMS 2010, Salt Lake City, UT, USA, 23rd – 27th May 2010, Mark Bayliss, Joseph Simpkins
The analysis of data supporting corporate compound library management, synthesis and medicinal chemistry support relies on LC/MS/UV/ELSD/CLND/CAD) as its primary means of substance confirmation. Confirmation being defined here as the presence of the substance, its purity (%Area of some chosen detector stream typically UV) and in some cases an empirical concentration calculation using CLND, ELSD or CAD. Our perception after performing millions of sample analyses is that we had to manually review more results and make more modifications than we felt was time efficient. Our greatest challenges were baseline determination inaccuracies, poor signal differentiation in the MS for weakly ionizing compounds, and poor assessment of adducts. Our challenge was to find a way to quantify these aspects and evaluate solutions. Starting with baseline accuracy, we investigated how a number of different baseline algorithms affected the performance of the final result by implementing the ability to capture that peak results immediately after processing and then after chemist review. Performing a difference analysis and then evaluating these results allowed us to investigate how accurately the different algorithms performed in routine usage and how adaptable they were to the typical problems commonly present in the different detector streams. The focus for our presentation will be on the accuracy of baselining in the determination of Area% which is used heavily as a determinant of implied substance purity. Holistically we can use the concept that the accuracy of processing can be normalized to: “The calculated difference between a processed result and the final result following review by a trained scientist” This approach has practical value as it can allow us to scientifically determine if the method that was used to process the data produces results that are statistically within the normal operating range that we would expect, based on similar model studies. In our preliminary studies we have found that baseline algorithms are certainly not all the same in their performance and that careful choice of baseline is extremely important if quality of %Area calculation is a key criterion in your analysis. Our presentation focuses on the comparative analysis of a number of different baseline algorithms and their accuracy in application to LC/MS/UV/ELSD/CLND (Or CAD) data that is typically used by scientists support compound validation.