Quantitative Systems Pharmacology and Machine Learning: A Match Made in Heaven or Hell?

Area of Science:

• Pharmacology
• Computational Biology
• Data Science

Background:

• Pharmaceutical development generates data across multiple scales, from subcellular to patient cohorts.
• Effective decision-making requires integrating diverse data types and knowledge.
• Machine learning (ML) and mechanistic modeling are key computational approaches.

Purpose of the Study:

• To outline the application of ML and quantitative systems pharmacology (QSP) models in pharmaceutical development.
• To guide the optimal integration of ML and QSP at various pipeline stages.
• To highlight the interplay between data-driven ML and knowledge-driven QSP.

Main Methods:

• Reviewing the application of ML and QSP models individually and in tandem.
• Discussing the role of sensitivity and identifiability analyses in QSP model development.
• Examining how ML can inform mechanistic model development and vice versa.

Main Results:

• ML and QSP models can be applied synergistically throughout the drug discovery and development pipeline.
• Discerning between available data and prior knowledge is crucial for model selection.
• Sensitivity and identifiability analyses of QSP models can guide experimental design for ML.

Conclusions:

• The combined application of ML and QSP models enhances decision-making in pharmaceutical development.
• The choice between ML and QSP depends on the availability of data and prior knowledge.
• Future research should consider the dynamic interplay between these approaches amidst continuous data acquisition.