Update batch flowjo3/20/2023 Specific features (histogram peaks or valleys) of manually gated or semi-manually selected data are used to modify samples to match a reference sample. A per-population “local” approach 18 builds upon fdaNorm, tightly integrating local (subpopulation specific) intensity normalization with the gating process. The registration (fluorescence normalization) programs fdaNorm and gaussNorm 17 normalize one channel at a time but require pre-gating of a subpopulation and do not address multidimensional linkages between biological subpopulations. Of these three classes of approach, registration has the advantage that it is a preprocessing step that leaves open the possibility of subsequently analyzing the cell subpopulations by any of the wealth of flow analysis programs now available. (3) Registration (data alignment) approaches are normally preprocessing steps to move the data directly (i.e., register) to improve alignment across samples 17, 18. Two of these methods 15, 16 try to mitigate adverse effects of batch variation during the cluster matching process using a random effects model during the matching process. (2) Cluster matching approaches attempt to match clusters between individually clustered samples to achieve a 1-to-1 or 1-to- N mapping of clusters with similar characteristics across samples 8, 9, 10, 11, 12, 13, 14, 15, 16. However, as batch variation increases, cells are assigned more frequently to the wrong clusters. These methods work well if batch variation is smaller than the biological variation of interest, for example, our SWIFT algorithm 3 was able to handle substantial variations between analysis centers 7. Individual samples can then be assigned to the template model. ![]() The strategies used to address inter-sample variation in such methods broadly fall into three categories: (1) Template-based methods generate models from selected or pooled samples and store all the relevant parameters (e.g., centroids, shapes, proportions, etc.) 1, 2, 3, 4, 5, 6. There are now many excellent automated methods to identify subpopulations in flow cytometry samples. However, these approaches are limited to two-dimensional (2D) gates, require a priori identification of gates, and fail in the presence of large shifts in subpopulation location. Some manual gating software packages offer auto-positioning gates to adjust for batch effects. Manual gating analysis can be selectively adjusted to deal with some batch or individual variability, but the process is time-consuming and subjective. Thus methods to reduce variability should ideally be objective, yet selective, for certain types of variability. In any study, only one or a few of these sources of variation will be the target of investigation of the study-the others should be minimized so that the target variation (e.g., therapy-induced changes) can be analyzed clearly. Fluorescence intensities may also be affected by biological variations, including genetics, environment, disease, age, gender, lifestyle, therapy, or microbiome. More complex changes in multiple channels are induced by variables that affect cell health and viability, e.g., shipping, cell handling, thawing, processing, and operator variability. The fluorescence of positively stained cells can be affected by staining protocol variations, antibody batches, or reagent instability. Day-to-day (batch) variations in global channel values can be caused by cytometer settings (e.g., photomultiplier tube (PMT) voltages, laser power, or different cytometers). Several types of variability contribute to the changes in marker intensity (using fluorescence or mass labels) that are inherent in flow cytometry. swiftReg outputs registered datasets as standard .FCS files to facilitate further analysis by other tools. swiftReg selectively reduces batch variation, enhancing detection of biological differences. Batch variation is addressed by registering batch control or consensus samples, and applying the resulting shifts to individual samples. Subpopulations are aligned between samples by displacing cell parameter values according to registration vectors derived from independent or locally-averaged cluster shifts. A high-resolution cluster map representing the multidimensional data is generated using the SWIFT algorithm, and shifts in cluster positions between samples are measured. We now describe swiftReg, an automated method that reduces undesired sources of variability between samples and particularly between batches. Each variation type requires a different correction strategy, and their unknown contributions to overall variability hinder automated correction. Biological differences of interest in large, high-dimensional flow cytometry datasets are often obscured by undesired variations caused by differences in cytometers, reagents, or operators.
0 Comments
Leave a Reply.AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |