Daniel Holmes (Presenter)
University of British Columbia
Bio: Dan Holmes did his undergraduate degree at the University of Toronto in Chemical Physics and then attended the University of British Columbia Medical School, graduating in 2001. He went on to complete his residency in Medical Biochemistry in 2006 is the Division Head of Clinical Chemistry at St. Paul's Hospital Department of Pathology and Laboratory Medicine in Vancouver Canada and holds the rank of Clinical Associate Professor at UBC. Research interests include Endocrine Hypertension, Endocrine Assay Development and Laboratory Medicine Statistics and Bioinformatics.
At present, LC-MS/MS vendors do not have upload/download capabilities to laboratory information systems (LIS). As a result, though tremendous energy is often spent in development of accurate, precise and robust LC-MS/MS methods, the results themselves are manually transcribed. This work is both demoralizing and error-prone. A simple approach to instrument datafile processing is presented using the R statistical programming language. Identification of patient results, calibrator, and QC is shown. Handling of non-numeric results is discussed. Potential hazards will be reviewed. Suppression of ion-ratio failures and the appending of interpretive comments are demonstrated. Source code will be available for free use or modification.
Mass spectrometry assay developers expend a great deal of energy preparing and validating methods showing excellent performance characteristics but often fail to consider that the results they generate will have to be manually transcribed into the laboratory information system (LIS). This transcription work is not only tedious and demoralizing, it is time-consuming and fraught with hazards. Most LISs offer the ability to perform flat-file (i.e. ASCII file) upload, provided that the file is appropriately pre-processed and formatted. Using the R statistical programming language, the instrument output file can be parsed for all relevant data in the file: results, QC, calibrators, qualifier/quantifier, and IS counts. Results data can easily be purged of non-numeric values and appropriately formatted for LIS upload, typically via SSH.
Datafiles from the SCIEX API 5000 and 5500 systems are parsed. A robust approach for handling the “ragged right” nature of the file is discussed. Non-numerical results are identified and replaced with LIS compatible values as appropriate. Analyte result columns are identified and regular expressions are used to determine QC, calibrator, and patient specimens. Ion ratios and failures can be determined and results censored. File formatting appropriate to LIS (in this case SunQuest) is undertaken and output file is written, upload-ready, with timestamp. Auto-archiving of raw file and processed file is demonstrated. Data are uploaded via SSH to SunQuest and reviewed for release in LARS or Autofiled as desired. Method has also been applied to the IDS iSYS immunoassay analyzer. Middleware interpretation algorithms have also be implemented since canned interpretive comments can be appended to parsed results.
This R middleware has been in production for over three years with no issues. Bench technologist hand-transcription of the LC-MS/MS results has ceased and approximately 50,000 results have been processed and uploaded. Source code will be made available for free use or modification.
An algorithmic pipeline for the automatic processing of LC-MS/MS results to the LIS using the R programming language is presented including detailed description of “bullet-proofing”. This obviates the need for manual transcription and avoids the use of expensive middleware solutions, which generally have more features than necessary. Source code will be made available.
References & Acknowledgements:
IP Royalty: no
|Planning to mention or discuss specific products or technology of the company(ies) listed above:||