A rapid, high-throughput, viral infectivity assay using automated brightfield microscopy with machine learning

Area of Science:

• Virology
• Biotechnology
• Artificial Intelligence

Background:

• Viral infectivity assays are crucial for developing vaccines, antivirals, and biologicals.
• Conventional assays are time-consuming (2-7 days) and require extensive manual processing.
• There is a need for faster, more efficient methods to quantify viral infectivity.

Purpose of the Study:

• To introduce an automated viral infectivity assay (AVIA) that significantly reduces assay time.
• To demonstrate the capability of AVIA to quantify viral infection phenotypes rapidly using AI.
• To compare the performance of AVIA with traditional infectivity assays.

Main Methods:

• Developed AVIA using convolutional neural networks (CNNs) and high-throughput brightfield microscopy on 96-well plates.
• Trained CNN models on diverse viruses including HIV, influenza, coronaviruses, vaccinia, poliovirus, and adenoviruses.
• Quantified infection phenotypes within hours, before manual visibility, without sample preparation.

Main Results:

• AVIA quantifies viral infection phenotypes within hours, drastically reducing assay time.
• CNN models demonstrated sensitivity comparable to or exceeding traditional plaque or TCID50 assays.
• Achieved a precision of approximately 10%, outperforming conventional infectivity assays.

Conclusions:

• AVIA offers a rapid, automated, and sensitive method for viral infectivity quantification.
• The AI-driven approach has broad-spectrum potential for virus characterization and identification.
• This technology can accelerate vaccine development, antiviral therapy research, and biological manufacturing.