Fig. 2

Identifying barcodes and splitting reads from Macosko et al., species mixing experiment. a Circular paths were identified in the circular barcode k-mer graph from a published Drop-seq dataset. The distribution of circular path weights versus path rank clearly shows an inflection point. Paths with weight higher than this inflection point are deemed to be true barcodes. b This inflection point can be identified as a local maximum in the first derivative of the path-weight distribution. A Savitskiy Golay filter facilitates in this identification by smoothing the data. c Reads were grouped into cells by assigning them to to thresholded paths based on k-mer compatibility alone. This assignment results in a flat distribution in the number of pseudoalignments per cell. d Reads that were split based on barcode k-mer compatibility alone also segregate by their number of pseudoalignments to different transcriptiomes. This indicates that assigning reads based on k-mer compatibility produces distinct and biologically relevant groupings