Multivariate Data Analysis (MVDA) | Onsite Course

Course Objective

The objective of this course is to guide the participants through their journey from a process data set to a multivariate monitoring model based on their process data. Through this course, multivariate data tables are translated into interpretable charts and plots, which simplifies the process monitoring and analysis of process deviations. Multivariate technology is the science of separating the signal from the noise in data containing many variables and presenting the results in a simple graphical format. This allows users to quickly go from a complicated table of numbers to a simple plot of the essentials. The key to unlocking the information in any data set lies in the correlations among the variables, not in the variables themselves.

Who Should Participate?

The course is intended for researchers, scientists, and engineers from all sectors of biopharma with a lot of process understanding, but with no or limited statistical background. Typical applications include product development, process improvement/optimization, deviation investigations, and scale-up of processes. Prior knowledge of statistics is beneficial but not needed.

Day 1, morning session

• Introduction to Sartorius Data Analytics
• Introduction to Multivariate Data Analysis (MVDA)
• The need for multivariate data analysis
• Customer use cases
• Introduction to multivariate modelling
• Introduction to PCA
• Principal Component Analysis (PCA) for overview of data tables
• What is Principal Component Analysis (PCA)?
• What is a score and what is a loading in PCA?
• Running the Workset wizard
• Software click-along demo and exercise

Day 1, afternoon session

• PCA Metrics and Diagnostics
• Outlier detection
• Scores plot, Hotelling’s T2 and DModX
• Loadings and contributions of variables
• Model Diagnostics for PCA, R2 and Q2
• Number of Components for a PCA model
• Variable scaling
• Variable centering
• Geometrical interpretation
• Model evaluation
• How does PCA bring value?
• PCA application: Counterfeit modelling
• PCA application: Raw material characterization
• PCA application: Process monitoring of a continuous process
• Software click-along demo and exercise

Day 2, morning session

Introduction to OPLS, Visualization

• Introduction to the MVDA regression methods OPLS® and PLS
• Applications of multivariate Regression
• Prediction of responses Y from parameters X
• Geometrical Representation of OPLS®
• Workset Wizard for OPLS
• Software click-along demo and exercise
• Model diagnostics and validation of an OPLS® model
• Scores plot, Hotelling’s T2 and DModX
• Loadings and contributions of variables for an OPLS® model
• Model Diagnostics for OPLS®, R2 and Q2
• Number of Components for an OPLS® model
• Software click-along demo and exercise

Day 2, afternoon session

• Visualization, documentation and reporting
• OPLS® Applications – how does OPLS® bring value?
• OPLS® Application: Discriminant analysis, What is the difference?
• OPLS® Application: Multivariate calibration
• OPLS® Application: Batch process modelling
• SIMCA® modelling for Ambr ® cell culture development
• Exercises
• Course Summary. What are the next steps?

Course Material and Software

The electronic course material (course slides and exercises) and the SIMCA® course license will be emailed two weeks before the course starts. SIMCA® needs admin rights to be installed, so please get in touch with your organization’s IT as soon as possible. Sartorius cannot help you to install SIMCA®. For onsite courses, we recommend that each participant has access to the software on their own PC and that a projector is made available for the course Instructor.

Preparation

We recommend listening to MVDA 1 - Multivariate Data Analysis (MVDA) for the Beginner (Link) and MVDA 2 - OPLS® in process modeling (Link) before the course starts. If you have any questions, please do not hesitate to send an email to: umetrics_academy@sartorius.com 

Cost and Conditions

The 2-day onsite course fee is 8360€ for up to 8 participants at the client location. A fee of €306 is added for each additional participant per day. Consultant travel and lodging is not included in this price.

Cancellations received later than two weeks before the course will not be refunded. For courses cancelled more than two weeks before the course starts, Sartorius Stedim Data Analytics AB will retain 10% of the course fee to cover administrative costs, and the rest of the amount will be refunded.

The registering company can substitute course participant(s) as long as Sartorius Stedim Data Analytics AB is notified before course commences.