Question: Loupe Browser provides three scaling options, i.e. Linear, Log2, and LogNorm. Which one should I choose when profiling gene(s) of interest in violin or t-SNE/UMAP plots?
Answer: In Loupe Browser, when profiling gene(s) of interest, three scaling options are available, i.e. Linear, Log2, and LogNorm. Linear provides raw UMI counts, Log2 renders log2-transformed raw UMI counts, and LogNorm offers normalized UMI counts (normalized by detected RNA content, same as the methods used in Seurat and Scanpy). Which option to use depends on the purpose of data visualization.
If data visualization is for explorative purposes, it doesn’t matter which scale value to choose. It should be noted that the value range for Linear (raw UMI counts) could be skewed by cells/spots with very high raw UMI counts for a given gene. Thus, Log2 and LogNorm values are recommended for easier visualization. However, if quantitative interpretation is needed, such as comparing gene expression levels between clusters, LogNorm is recommended because it normalizes the detected RNA content in each cell/spot.
Loupe Browser also supports the visualization of multiple genes (with Attribute parameters: Feature Max, Feature Min, Feature Sum, Feature Avg, or Total UMI counts). For explorative purposes, any of the three scaling options (Linear, Log2, or LogNorm) could be used. However, if you would like to compare expression levels of multiple genes across cells, you should use the normalized value (LogNorm) and, meanwhile, consider the fact that some genes are intrinsically more abundant than some other genes. Additional data standardization (not included in Loupe Browser) may be needed for proper comparisons with multiple genes.