Bioprocess Data Analytics and Machine Learning

• Richard D. Braatz
• Brian Anthony
• Seongkyu Yoon

For more information please read the Bioprocess Data Analytic and Machine Learning schedule (pdf).

Description

This course may be taken individually or as part of the Professional Certificate Program In Machine Learning & Artificial Intelligence or the Professional Certificate Program in Biotechnology & Life Sciences.

Biotherapeutics has improved the lives of millions of patients around the world. In the past few years, major advances in biomanufacturing analytics, analytical technology, and machine learning have deepened understanding of process operations and product quality in this important arena.

Organizations on the leading edge of bioprocess data analytics have already seen dramatic improvements in pharmaceutical batch optimization, manufacturing scalability, and regulatory efficiency.

To help you take advantage of these revolutionary developments—and drive breakthroughs of your own—MIT Professional Education is pleased to introduce Bioprocess Data Analytics and Machine Learning. In this intensive, three-day course, you’ll gain:

• A greater understanding of how bioprocess data analytics can be applied to develop and improve biotherapeutic manufacturing
• Insight into important advances in data analytics, machine learning methods, and software that provide new ways to build models, diagnose problems, and make informed decisions
• An introduction to new sensor technologies, including spectral imaging and real-time color video, and the major classes of data analytics and machine learning methods used in bioprocess operations
• Tools to systematically interrogate the data to ascertain specific characteristics needed to select among the best-in-class data analytics methods

With the guidance of academic and industry experts, you’ll discover transformative ways to apply data analytics—and avoid the most common pitfalls that arise when analyzing bioprocess data. By the end of the course, you’ll have an understanding of the best practices needed to translate biopharmaceutical manufacturing data into reliable models and better decisions. Simply put, you’ll be able to select the right methods, improve accuracy and effectiveness, and save time and money.

Participant Takeaways

• Apply new sensor technologies relevant to biopharmaceutical manufacturing processes, such as spectral imaging and real-time color video
• Understand major classes of data analytics and machine learning methods relevant to bioprocess operations
• Systematically interrogate bioprocess data to ascertain characteristics (such as nonlinearity, multicollinearity, and dynamics)
• Select among the best-in-class data analytics methods based on the objective and data characteristics
• Summarize ways to combine data-driven models with mechanistic understanding
• Avoid common pitfalls when analyzing bioprocess data

Who Should Attend

• Bioprocess Data Analytics and Machine Learning is designed for scientists and engineers in the biopharma industry who want to take their skills—and their careers—to the next level. In particular, this course is well suited to individuals with job titles such as Data Scientist, Senior Research Scientist, and Bioprocess Engineer. Additionally, workshop participants should have some experience with analyzing experimental data.

Requirements

• Participants should have some experience in data analytics and bioprocesses. In addition, participants must be professionals with experience working in the pharmaceutical or biopharmaceutical industry.

Program Outline

Classes will run on the following schedule:

• Day One: 9 a.m. – 7 p.m.
• Day Two: 9 a.m. – 6 p.m.
• Day Three: 9 a.m. – 6:30 p.m.

Day One

• 9 to 10:30 a.m. – Braatz/Anthony
  Introduction to bioprocess data analytics: opportunities, types of data analytics problems; supervised, unsupervised, and partially supervised learning; data visualization (software tools using data from bioprocess time series, Raman spectroscopy, spectral imaging, LC-MS)
  Readings/Assignments: Lecture notes
• 10:30 to 11 a.m. – Coffee and networking
• 11 a.m. to 12:30 p.m. – Braatz
  Regression: least squares, response surface methodology, ridge regression, lasso, elastic net, cross-validation, residual analysis, outlier detection, uncertainty quantification, optimal experimental design, feature engineering, nonlinearities, systematic method selection, monoclonal antibody manufacturing industrial case study
  Readings/Assignments: Lecture notes
• 12:30 to 1:30 p.m. – Lunch and networking
• 1:30 to 2:30 p.m. – Braatz
  Tour of biopharmaceutical manufacturing labs at MIT, which includes fully automated and instrumented end-to-end continuous CHO-based monoclonal antibody production
  Readings/Assignments: Lecture notes
• 2:30 to 3:30 p.m. – Braatz
  Tips and traps: correlation vs. causation, inferences drawn from matching statistics, effects of feedback loops, selection among too many models, studying features for biological/ chemical/physical reasonableness, Google Flu Trends
  Readings/Assignments: Lecture notes
• 3:30 to 5 p.m. – Anthony/Braatz
  Hands-on activities in visualizing and analyzing real biopharmaceutical process datasets
  Readings/Assignments: Students use their own laptops
• 5 to 7 p.m. – Course reception/networking

Day Two

• 9 to 10:30 a.m. – Yoon / Anthony
  Time series analysis: low-, high-, and band-pass filters; oscillatory data; Fourier transforms; ARMA; ARMAX; noise; AD convertors; sampling; aliasing; statistical process control (aka control charts), UV absorption for protein concentration measurement case study
  Readings/Assignments: Lecture notes
• 10:30 to 11 a.m. – Coffee and networking
• 11 a.m. to 12:30 p.m. – Yoon
  Latent variable methods I: PCA, multivariable statistical process control, spectral sensor calibration, PCR, PLS, spectral data artifacts, Raman spectroscopy case study
  Readings/Assignments: Lecture notes
• 12:30 to 1:30 p.m. – Lunch and networking
• 1:30 to 3 p.m. – Yoon
  Latent variable methods II: CCA, FDA, missing data, sparse models, Raman and near-infrared case studies
  Readings/Assignments: Lecture notes
• 3 to 3:30 p.m. – Coffee and networking
• 3:30 to 4:30 p.m. – Yoon/Braatz/Anthony
  Hands-on activities in visualizing and analyzing real biopharmaceutical process datasets
  Readings/Assignments: Students use their own laptops
• 4:30 to 5:30 p.m. – Yoon/Braatz/Anthony
  Open discussions with the instructors
• 5:30 to 6 p.m. – Coffee and networking

Day Three

• 9 to 10:30 a.m. – Anthony/Braatz
  Big data analytics: real-time video, thermal imaging, tensor data, hyperspectral imaging, LC-MS, multiway methods, multilinear subspace learning, lyophilization case studies
  Readings/Assignments: Lecture notes
• 10:30 to 11 a.m. – Coffee and networking
• 11 a.m. to 12:30 p.m. – Anthony
  Nonlinear analytics: support vector machines; random forests; comparisons to FDA and k-nearest neighbor classification; feature engineering revisited; kernel methods; hybrid models; autoassociative and recursive neural networks
  Readings/Assignments: Lecture notes
• 12:30 to 2:30 p.m. – Course lunch and networking
• 2:30 to 4 p.m. – Anthony
  What can bioprocess analytics learn from other industries?
  Readings/Assignments: Lecture notes
• 4 to 4:30 p.m. – Coffee and networking
• 4:30 to 5:30 p.m. – Anthony / Braatz
  Hands-on activities in visualizing and analyzing real biopharmaceutical process datasets
  Readings/Assignments: Students use their own laptops
• 5:30 to 6:30 p.m. – Anthony/ Braatz
  Closing reception

Registration

Registration closed.