High-Throughput Virus Quantification by Cytopathic Effect Area

Abstract

Traditional infectivity-based virus quantification methods, such as plaque assays and TCID50, face significant limitations when scaled to modern high-throughput formats like 384- or 1536-well plates. Although recent advances in computer vision and machine learning have improved aspects of analysis, these assays remain constrained by the surface area and manual labor requirements of legacy protocols. In this study, I present a scalable alternative that quantifies infectious virions by measuring virus-induced cytopathic effect (CPE) via cell lysis and detachment, using whole-well image thresholding. This CPE area assay enables accurate, automated quantification compatible with high-throughput screening formats, offering greater efficiency than traditional viral infectivity methods. In an attempt to enhance prediction accuracy, dynamic range, and reduce reliance on standard curves for the CPE area assay, I evaluated two deep learning based image analysis models: Siamese Neural Networks (SNNs) and Vision Transformers (ViTs). Although both models showed promise, particularly the ViT in generalization across viral datasets, the threshold-based CPE area analysis remained the most accurate and generalized method.