Our Solution

The “chemometrical calibration” problem

Multivariate measurement instruments, e.g., infrared or Raman spectrometers, need to be calibrated to the particular measurement application at hand before they can output user-relevant information. When the signal from the analyte of interest does not clearly stand out against the background, univariate calibration is not sufficient and multivariate calibration is needed. The “statistical” methods used so far for multivariate calibration, mostly PLS, enjoy a less than shiny reputation in industry. While generally appropriate for data-exploration type tasks, these methods are generally inappropriate for calibration. The statistical approach makes the process of calibration expensive, often more expensive than the instrument hardware itself and, no matter how high the effort, the quality of the result remains questionable. Calibration is therefore currently the main technical hurdle and de facto bottleneck to the successful deployment of multivariate instruments. This can have significant economic consequences, e.g., in a manufacturing company when a desired process control measurement can not be realized.

The solution

The solution is a new method of multivariate calibration, see Refs. [1-3], which practitioners in industry have named Science Based Calibration (SBC). SBC simplifies the calibration process, improves the quality of the results, and makes the results interpretable. Compared to prior methods, the main advantages are:

• The need for lab-reference values is virtually eliminated from calibration. The cost of calibration can therefore be significantly reduced compared to today’s standard practice of PLS or PCR, often by as much as 80%.
• The need to artificially upset an industrial process in order to collect on-line calibration standards that vary over a certain range is eliminated. Rather, a smoothly running process with a minimum of analyte variation is sufficient.
• The quality of calibration is improved, in particular, robustness
• Specificity of response can be proven straightforwardly and from first principles. This is a fundamental advantage.
• New and improved possibilities become available to tackle calibration transfer
• Spectrometer hardware performance figures can be directly translated into user-relevant measures of output accuracy. This simplifies the setting of hardware specifications, which in turn enables significant reductions in R&D time and expense of development programs for new, application-specific instruments.
• The calibration process itself becomes fully transparent to the user, or in the words of the pharmaceutical industry, science-based. Results can thus be easily communicated and assessed by third-parties

All this is achieved, in a nutshell, by replacing statistics with science.

An easy-to-read summary of the main technical points is published in a two-part trade journal article [4]. The first part of the article outlines the theory, history, relationship to other methods, and advantages of SBC. The second part of the article addresses applicability, limitations, and the important issue of “multivariate specificity.” The correct definition of the latter has not been available previously, compare [5], but is straightforward when using SBC theory and nomenclature.

Our solution

Science Based Calibration is protected by US Patent 6,629,041. MTT Multantiv owns the rights to this patent, including the right to license. SBC can provide significant advantages to commercial enterprises.

Target customers for Science Based Calibration include user companies, i.e., companies using multivariate instruments in their operations and having to calibrate them (e.g., manufacturing companies in the chemical or pharmaceutical industry) and manufacturers of dedicated instruments, i.e. “pre-calibrated” multivariate measurement instruments (e.g., biomedical device companies or automation suppliers).

MTT Multantiv can help you to drastically simplify the process of multivariate calibration. Tasks related to calibration, e.g., certain R&D processes or validation activities can often be simplified as well.

Commercialization of SBC has started. We are interested to provide licenses to the SBC solution and to discuss potential co-operations. Please contact us for details.

A free license is available for all non-commercial uses for those who work in academia, a government agency, a standard setting body, or in any other kind of not-for-profit organization where the content of the work is not related to any commercial activity. Please contact us for details.

An evaluation license is also available, including the option of a short-term consulting service by MTT Multantiv. Days rather than weeks of time can be sufficient to solve an example problem and train your staff in the process. Please contact us for details.

References

1. R. Marbach, On Wiener Filtering and the Physics Behind Statistical Modelling, J. Biomed. Optics 7, 130-147 (2002)
2. R. Marbach, Methods to significantly reduce the calibration cost of multichannel measurement instruments, US Pat. No. 6,629, 041, 30 Sep. 2003
3. R. Marbach, A New Method for Multivariate Calibration, J. Near Infrared Spectrosc. 13, 241-254 (2005)
4. R. Marbach, Multivariate calibration: A science-based method – Part 1, Pharmaceutical Manufacturing 6(1), 42 – 47, January 2007 (on the advantages). Part 2 (on specificity and the need for action) in Pharmaceutical Manufacturing 6(2), 44 – 47, February 2007. These articles are also available at www.pharmamanufacturing.com
5. A.C. Olivieri, N.M. Faber, J.Ferré, R. Boqué, J.H. Kalivas, H. Mark, Uncertainty Estimation and Figures of Merit for Multivariate Calibration (IUPAC Technical Report), Pure Appl. Chem. 78(3), 633–661, 2006