Welcome!

A recent report of my research on FDA website!

Artificial Intelligence/Machine Learning Assisted Image Analysis for Characterizing Biotherapeutics

How AI/ML Can Help Characterize Biotherapeutics

Biotherapeutics are drugs derived from living organisms, such as proteins, antibodies, or vaccines. They have revolutionized the treatment of various diseases, such as cancer, autoimmune disorders, and infections. However, biotherapeutics are also complex and sensitive molecules that can form unwanted particles or aggregates under different stress conditions. These particles can affect the quality, safety, and efficacy of the drug product, and may trigger adverse immune responses in patients.

To monitor and characterize these particles, scientists use various analytical techniques, such as flow imaging microscopy (FIM). FIM can capture large collections of images of individual particles in a sample, providing rich information about their size, shape, and composition. However, extracting and analyzing this information manually is time-consuming and subjective. Moreover, current methods rely on human-defined features, such as aspect ratio or compactness, which may not capture the full complexity and diversity of the particles.

This is where artificial intelligence/machine learning (AI/ML) can help. AI/ML is a branch of computer science that enables machines to learn from data and perform tasks that normally require human intelligence. One of the applications of AI/ML is image analysis, which involves processing and understanding digital images. In particular, convolutional neural networks (CNNs) are a type of AI/ML model that can automatically extract and learn complex features from images, such as edges, shapes, textures, or patterns.

Scientists at the FDA, in collaboration with external partners, have applied CNNs to analyze FIM images of particles in biotherapeutics. They have shown that CNNs can classify particles with high accuracy and efficiency, and reveal novel features that are not detected by conventional methods. For example, CNNs can distinguish between different types of protein aggregates, such as amorphous or fibrillar, or identify particles that are specific to certain stress conditions, such as shaking or freezing. These features can serve as fingerprints for the quality and stability of the drug product, and help identify the root causes of particle formation.

The use of AI/ML for biotherapeutics evaluation is an emerging and promising field that can enhance the understanding and characterization of these complex products. AI/ML can also complement and augment the existing analytical methods, and provide new insights and perspectives for the regulatory science of biotherapeutics. The FDA is committed to advancing and applying AI/ML to support the development and evaluation of safe and effective biotherapeutics for the benefit of public health.