By Diagnostics World Staff
August 5, 2025 | Researchers at Virginia Commonwealth University have developed an artificial intelligence tool that could revolutionize cancer diagnostics by rapidly identifying distinct cell types in tissue biopsies, potentially transforming how clinicians make treatment decisions and predict patient responses to therapy.
The AI algorithm, called TACIT (threshold-based assignment of cell types from multiplexed imaging data), represents a significant advancement in spatial biology diagnostics. In a recent study published in Nature Communications, TACIT successfully distinguished 51 different cell types across nearly five million cells, outperforming existing methods in both accuracy and scalability.
Dr. Kevin Matthew Byrd, associate research member at VCU Massey Comprehensive Cancer Center, explains that while pathologists have been diagnosing diseases based on spatial tissue context since 1875, TACIT can now analyze multiplexed slides in minutes rather than the 24-48 hours or even weeks currently required for complex assays.
Clinical Diagnostic Applications
The platform-agnostic algorithm addresses a critical diagnostic challenge by analyzing high-plex imaging data to reveal dozens or hundreds of molecular markers simultaneously within individual cells. This comprehensive cellular mapping creates what Byrd describes as "a snapshot of the disease or tissue in time," providing unprecedented diagnostic resolution.
TACIT's diagnostic power extends beyond traditional imaging approaches by combining slide analysis with transfer proteomics. This dual-analysis capability is particularly valuable for predicting patient responses to immuno-oncology drugs like Keytruda, where researchers have found that RNA expression often doesn't match protein levels—a critical factor in determining treatment efficacy.
Solving Diagnostic Complexity
The algorithm tackles what researchers call the "curse of dimensionality"—the difficulty of finding meaningful diagnostic patterns in high-dimensional biological data. Dr. Jinze Liu, co-developer and professor of biostatistics at VCU, explains that TACIT uses computational methods to focus on relevant cellular features that distinguish cell types, achieving robust diagnostic signals even in complex tissue environments. Unlike current graph-based clustering methods that require extensive human interpretation, TACIT operates as a self-learning algorithm that autonomously improves its diagnostic accuracy over time without external input.
Future Diagnostic Capabilities
The diagnostic potential extends beyond current applications as the Human Cell Atlas project continues mapping the estimated couple thousand cell types that comprise the human body. With close to 150 million single cells profiled to date, TACIT will leverage this expanding database to enable more sophisticated patient sub-stratification based on tumor microenvironments.
The algorithm could eventually provide clinicians with real-time insights into tumor growth, metastasis potential, treatment resistance mechanisms, and immune system status—all critical factors in diagnostic decision-making and treatment planning.