Software  ›   visualization

# Technical Details

Downsampling
Copy number and confidence values are called at 20kb resolution. Bins are downsampled based on the current zoom level and window size. The downsampled copy numbers for a region are computed by averaging all 20kb bins. If greater than half the bins are unmappable, the downsampled bin is then marked as unmappable.

Tree Traversal
The tree displays a subset of nodes in the tree computed by the pipeline. The nodes are selected using a greedy algorithm that starts at the current root and expands the node with the greatest number of leaves until it reaches the current number of nodes, chosen through the Settings in the Navigation panel. Loupe scDNA Browser shows a subset of nodes in order to scale to datasets with thousands of single cells.

Copy Number Color Scale
Copy numbers are discretized from 0-6 by ones, then by twos from 8-16. The bins are as follows: [0, 0.5), [0.5, 1.5), [1.5, 2.5), ..., [5.5, 7), [7, 9), [9, 11), ..., [16, ∞). Note that the bin for copy number as 6 is special. Its bin size is different from the others.

Exporting
Exported data will reflect the current view. If what is displayed is 16 nodes, the corresponding data for only 16 nodes will be exported. Downsampled values will be exported if large areas are being viewed. Bins which are unmapped will appear blank.

Measured read depth is quantized into approximate read depth in the .dloupe file, to save disk space. First, for each node, the array of GC bias-corrected read depths for each bin are converted to an "imputed copy numbers", which is the original read coverage per node at each bin, divided by average per-bin read coverage for that node, times the average ploidy of the node. The pipeline quantizes that imputed copy number into one of 252 int8 values, which are weighted in density toward the 0-8 region. The quantization can represent imputed copy numbers between 0 and 8 at .04 resolution, between 8 and 16 at .2 resolution, between 16 and 24 at whole integer resolution, and finally with a symbolic "imputed copy number > 24" value. Approximate read depths for a bin are computed back at runtime by dequantizing the stored integer value to the imputed copy number, dividing by the stored average ploidy of the node, and multiplying by the stored average read depth of the node. You can observe this approximation in action by hovering over the trace and noticing that the same read depth values tend to reoccur. It should be noted that the range of possible copy numbers is more precise than approximate read depth above 24.
For gene annotations to be searchable and viewable in Loupe files, a cellranger-dna pipeline must be run against a reference which contains a GTF file in the expected location (genes/gene_annotations.gtf.gz). For more details on creating such references, see how Reference Support. Loupe will embed gene features with gene_type or gene_biotype attributes with values protein_coding or pseudogene.