explorer is a tool for scientists investigating single cell rna seq datasets. The design of explorer proceeds from a small number of core principles:
- the tool, first and foremost, must produce views that are scientifically valid at all times
- the tool should be highly scaleable and handle exploration of millions of cells in the browser at interactive speeds, including interactive crossfiltering and dataframe subsetting
- the tool should be data dense and give scientists powerful views into data
- the tool should be powerfully expressive and optimized for the
nthday of use rather than the first day, in the spirit of enterprise tools, even if that requires training or onboarding- (think: photoshop, which is supported by a broad array of youtube tutorials, books and trainings)
- the tool should avoid duplicating data onscreen
- for example, if we render a categorical label on the left hand side bar, locate further information related to that label in place on the left hand sidebar, rather than rendering that data again.
- for a concrete example, see the relationship between this solution: chanzuckerberg/cellxgene#827 and this problem: chanzuckerberg/cellxgene#762
- the tool should enable users to rapidly test hypotheses on the application, which may require it to enable interactive recomputation of views into the data, for example:
- interactive differential expression
- interactive reprojection of umap
- the tool should enable a productive workflow between those who are computational and those who are not
- the tool should minimize extraneous use of color, leaving color to primary workflow actions like
compute differential expressionorcreate new categorical metadataorcreate new geneset. This leaves the color space tocolorbyactions, such ascolorby genesetandcolorby categorical field