High-throughput cytometry data processing

Algorithms for automated detection of sample signature changes in high-throughput cytometry-based drug-discovery assays

Cell-based assays and screening techniques are gaining popularity and becoming indispensable tools employed in drug discovery. The reason is that cell-based analysis methods provide better insight into complex cellular systems and allow the study of targets and pathways not accessible in simple single-point biochemical assays. Single-cell analysis techniques also allow for studying the heterogeneity of cellular populations, providing a data structure that contains information about the role of heterogeneity and its possible functional significance. However, one consequence of the complexity of studied cell systems is the complexity of the collected data. Therefore, feature extraction, data reduction, analysis, and mining represent a real and significant problem, limiting our ability to utilize modern single-cell techniques.

Two of the most critical technologies in cell-based screening are undoubtedly image-based automated microscopy tools and flow cytometry. Although the high-content screening community focused its attention on microscopy methods, FC tools have recently experienced a resurgence in popularity among screeners. This happened due to the introduction of modern autosampling devices (such as HyperCyt) providing high-throughput capabilities, the publication of highly multiplexed and multiparametric assays such as phosphoflow, and the active engagement of the informatics and machine-learning communities, leading to the development of new, highly sophisticated analytical techniques.

Yet FC measurements are still challenging to quantify in a manner amenable to automation and use in typical screening operations. It is clear that there is an unmet expectation for operator-independent, robust quantification methods that could produce robust readouts usable in the context of compound screening and drug discovery.

However, the solution to this problem requires designing a new conceptual framework for FC data analysis that radically departs from established traditions. Since the infancy of the method, FC data analysis has typically been performed in an exploratory fashion. However, high-throughput flow cytometry (HT-FC) techniques utilized to perform screening produce an extraordinary amount of multifactorial and multidimensional data which do not fit well into exploratory paradigms, either manual or automatic. Conceptually, multifactorial HT FC combines the cell-based utility of cytometry studies with the complex multifactorial format of single-point assays (See Figure 1 below). The vast quantity and high dimensionality of the data prohibit any use of traditional exploratory interactive techniques in a practical setting. In fact, owing to the nature of the problems studied with screening platforms, automation of the data-processing pipeline becomes the sine qua non for the successful deployment of any HT screening technology.

Page updated on 2022-12-07 17:24:35 -0500