One depth column (d|<region>: the window z, capped at +/- cap, NA -> 0) and one signed allelic
column (a|<region>) per eligible region of the given classes. Allelic z =
(nA - n / 2) / sqrt(n / 4 * (1 + (n - 1) * rho)) on the oriented votes, capped at +/- cap, 0 below
min_n molecules; the sign keeps which haplotype is in excess, so subclones that are imbalanced on
opposite haplotypes separate (e.g. two subclones with a loss of different chr21 homologues). Input whose orient is not in keep_orient
(unoriented block-phased none, or clone = oriented by known groups such as spatial regions, which would leak the
grouping into discovery) is first re-oriented by the all-cell leave-one-out pseudobulk (orient_blocks,
NULL groups) when votes, atoms, regions and cells are given.
Arguments
- feat_long
features long table (
cell_features()long)- classes
region classes used
- rho
beta-binomial overdispersion of the molecule votes
- cap
cap of |z|
- min_n
minimal molecules for an allelic value
- votes, atoms, regions, cells
optional: orient unoriented input (cells = names in vote index order)
- keep_orient
orientlabels used as given (chromosome-wide anchored phase, all-cell pseudobulk)