april.quinn-paquet@sciex.com

Member

Automatic outlier removal using SCIEX OS software

Advances in technology have resulted in a need for higher and faster sample throughput, with increasing number of measured analytes per sample – sometimes up into the 1000’s. A greater strain on necessary data analysis steps leads to a bottleneck of data processing and review for chromatographic peak integration, quality of calibration curve, and more. Analysts can spend hours just reviewing calibration curves for acceptability.

SCIEX OS-MQ software brings routine quantitation workflow to the next level for quicker turnaround time. The sleek, intuitive user interface overlies a sophisticated new peak integration algorithm and other new features, such as the automatic outlier removal for calibration curves, and new tools for calculating, flagging, and filtering results. These features combine to make data processing and review more streamlined and efficient.

The Automatic Outlier Removal feature allows the user to set criteria for the automatic removal of standard outliers from the calibration curve. The user controls what regression linearity, accuracy of standards, precision of replicates, and outlier tolerance are acceptable, and the algorithm then builds the calibration curve for each analyte to meet the defined criteria.

How does it work? The software iteratively surveys all data points to identify a starting range consisting of 4 consecutive points (minimum of 3) that provide the best linear regression and satisfy the user-specified rules for outlier removal. The algorithm calculates multiple regressions for all permutations of the starting points and considers ALL valid regressions that satisfy the user criteria and takes all of them through the expansion sequence. For all the valid starting ranges, the success of each expansion depends on the total number of used points, the range of the used levels, and the point with the worst absolute accuracy error in the regression before and after the expansion. The regression that spans the larges range and satisfies the user-specified criteria is the “winning” regression.

This new feature helps to alleviate the pain of manually establishing many calibration curves when working with many analytes varying in sensitivity and performance. In an example of 193 analytes with 9 calibrator levels acquired in triplicate, the processing took about 1 minute, compared to 180 minutes to manually review and approve all calibration curves for the same data set. This is a leap forward in time saving for data reviewers.

Improving routine quantitation using SCIEX OS-MQ Software, SCIEX technical note RUO-MKT-18-10333-A.

RUO-MKT-18-10333-A

Join the discussion

Comment below on this article and our team will answer your questions.

0 Comments

Submit a Comment