Punkst Modules¶
punkst provides several command-line tools for analyzing high resolution spatial (transcriptomics) data.
Available Modules¶
- pts2tiles: Group pixels to tiles for faster processing
- tiles2hex: Group pixels into non-overlapping hexagons for spot level analysis
- topic-model: Run LDA on the spot level data
- pixel-decode: Annotate each pixel with the top factors and their probabilities
- cooccurrence: Compute gene co-occurrence and/or extract marker genes from the co-occurrence matrix
- visualization: Visualize the pixel level analysis results
- DE tests: Differential expression tests
- tile-op: View and manipulate high resolution spatial inference output (merge, annotate, denoise, bin, etc.)
- pmtiles: Export PMTiles archives and build point / polygon MLT pyramids
Input Data Format¶
The input is a tsv file with the following columns: X, Y, feature, count. Whether the file contains headers or other columns is not relevant, as long as the above four columns are present.
-
X, Y coordinates can be either integer or float vlaues. (If your coordinates are integers and you would like to keep the original coordinates in the pixel level inference output, set
--coords-are-intinpunkst pixel-decode). The coordinates can be in arbitrary units, just make sure all scale/size related parameters you later provide are in the same unit. -
"feature" column contains the gene names (strings) or a nonnegative integer corresponding to the index in a feature list.
-
"count" is a nonnegative value for transcript counts (it does not have to be integers). If your data are from imaging platforms where each record is a single transcript, you might need to add this column manually with all
1.