Fig. 2: Cell type signatures inferred from cell-free DNA fragmentation are preserved at ultra-low sequencing depths.

A Tracks for sequencing coverage and window protection scores (WPS) are shown for sample “GC01” sequenced at 35-fold (35×) coverage and a randomly selected sample sequenced at 10-fold (10×) and <0.3-fold (<0.3×) coverage for the first 5 kilobases (kb) of the FCN1 gene. Per-base WPS was calculated by taking the number of fragments that span a 120 base pair (bp) window minus the number of fragments that have an endpoint within that same window centered at a given genomic coordinate. B The distance (bp) between the genomic coordinate of each WPS peak call in the set of healthy control samples and the closest WPS peak call in published sample “CH01” from Snyder et al. (2016) is plotted. Distances are grouped by sequencing coverage of the healthy control samples: 35-fold (n = 1), 10-fold (n = 139), and <0.3-fold (n = 90) coverage. The published call set in sample “CH01” was sequenced at 231-fold coverage (12.9 million peaks). C The distance (bp) between adjacent peak calls (i.e., inter-nucleosomal distance) is plotted genome-wide for all healthy control samples grouped by sequencing coverage. Ranked correlation values between fast Fourier transformed (FFT) WPS in the first 10 kb of 19,536 genes and the average expression of these genes in 456 cell types from the Tabula Sapiens database are grouped by D compartment and E tissue for samples sequenced at 35-fold (n = 1), 10-fold (n = 139), and <0.3-fold (n = 90) coverage. Ranked correlation values are stratified by cell type for the immune compartment F for all healthy control samples (n = 230). Boxplots indicate median, interquartile range (IQR), and whiskers for 1.5× IQR. Cell types were annotated by the original publication with four compartments (i.e., immune, epithelial, endothelial, or stromal) and originated from 24 different tissue types biopsied by the Tabula Sapiens consortium G Ranks for the highest (type I NK T cell) and lowest (erythrocyte) correlations are plotted for sample “GC01” when down-sampled H The macrophage rank distribution grouped by tissue-residency is plotted for all healthy control samples (n = 230). NK natural killer, Eff. effector, DC dendritic cell, β beta, α alpha. Source data are provided as a Source Data file.