Application of machine learning methods to pathogen safety evaluation in biological manufacturing processes

Abstract

The production of recombinant therapeutic proteins from animal or human cell lines entails the risk of endogenous viral contamination from cell substrates and adventitious agents from raw materials and environment. One of the approaches to control such potential viral contamination is to ensure the manufacturing process can adequately clear the potential viral contaminants. Viral clearance for production of human monoclonal antibodies is achieved by dedicated unit operations, such as low pH inactivation, viral filtration, and chromatographic separation. The process development of each viral clearance step for a new antibody production requires significant effort and resources invested in wet laboratory experiments for process characterization studies. Machine learning methods have the potential to help streamline the development and optimization of viral clearance unit operations for new therapeutic antibodies. The current work focuses on evaluating the usefulness of machine learning methods for process understanding and predictive modeling for viral clearance via a case study on low pH viral inactivation.

Keywords: biological manufacturing process; low pH viral inactivation; machine learning; monoclonal antibody; pathogen safety.

FIGURE 1

Schematic of purification process

FIGURE 2

For perfect class separation, there should not be any point in the band of width 2M. The red colored data points in the blue region and the blue colored data points in the red region are considered incorrectly classified. In case, a perfect separation is not possible, margin M is maximized by constraining the total distance of incorrectly classified data points from their respective margins

FIGURE 3

Decision tree for classification

FIGURE 4

Score plot provides an overview of performed experiments in a reduced dimensional space. Each data point in the score plot represents a single experiment. Score plot also helps in identifying groupings as shown by highlighted regions (Group‐1 and Group‐2)

FIGURE 5

Contribution plot enables the identification of process parameters responsible for the grouping of experimental runs

FIGURE 6

Loadings represent the relationship between the score vectors in a reduced dimensional space and process parameters

FIGURE 7

Confusion matrix for: (a) OPLS‐DA, (b) LR, (c) SVM, (d) DT, and (e) RF machine learning algorithms

FIGURE 8

Predictive loadings for OPLS‐DA classification model

FIGURE 9

A decision tree from the Random Forest classification model