( G) This classifier is then used for deconvoluting data from new experimental sets and assigning each event to a CNN class with a given probability. This will generate a trained classifier with CNN classes based on FDL clusters. A CNN classifier is trained using the images obtained from homeostasis, naïve or wild-type (WT) cells, and already organized in clusters in an unbiased way through the first part of our method. (2) If the goal is integrating experiments and comparing cell type abundance between them, the use of steps ( F) and ( G) is suggested. Compare sets of clusters coming from multiple experiments and multiple rounds of analysis can be challenging without pre-existing knowledge of cell types, clearly different morphologies or biomarkers that would allow to establish a unique correlation between clusters coming from different FDL graphs. This approach will produce a new set of clusters that will need to be reannotated. control), the steps described so far from ( A) to ( E) can be reapplied to the new dataset including a statistical analysis to compare cluster relative abundance. (1) If the goal is comparing samples belonging to the same experiment ( e.g., treatment vs. ( F) If new experiments are run and new data needs to be analyzed, two approaches can be taken. This heatmap shows feature similarities and differences between cells belonging to different clusters. ( E) Spearman’s correlation plot of feature values by clusters is one of the options available in Image3C for plotting integrated data. This step allows to evaluate the morphological homogeneity of the clusters, determine if the number of clusters is appropriate, and explore the phenotype/function of the cells based on visualization of individual channels. ( D) R integration in FCS Express Plus software allows the visualization of cell images by clusters or specifically selected with a gate. ( C) Images are clustered based on morphological and fluorescent feature values and visualized as a force-directed layout (FDL) graph where each dot represents one event. Samples that are outliers among replicates are also removed prior to the final normalization of the fluorescence intensities. R (or R studio) is used to calculate the correlation between features to allow to trim the features that are redundant with others. ( B) IDEAS software is used to open the raw images, compensate for correcting fluorescent spillover, subtract background, and quantify values for intrinsic morphological and fluorescent features. The samples are run on the ImageStream X Mark II, and 10,000 nucleated and focused events are saved for each sample as individual raw images. The signal can highlight specific cell components ( e.g., nuclei), metabolic cell states, or specific cell functions. ( A) A single cell suspension is prepared for image-based flow cytometric analyses. The cells can be labeled with any reagent working for the species of interest.
0 Comments
Leave a Reply. |
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |