Remove invariants during representative markers choice

[aathbt]

The representative markers are selected only on bimodality. It could be interesting to add another filter to remove markers with a low variance to avoid wasting time on markers that will probably not be used to characterize a cell population.

Former code:

vt = VarianceThreshold().fit(mat_subset)
invariants = markers[vt.variances_ <= np.max(vt.variances_[np.isin(markers, marker_order)])]
# markers_rep = np.asarray(markers_rep)[np.isin(markers_rep, invariants) == False]
markers_rep = np.asarray(markers_rep)[np.invert(np.isin(markers, invariants))]
markers_rep = markers[np.isin(markers, markers_rep)]

Use a threshold X depending on min/max values of expression? VarianceThreshold(threshold=X).fit_transform(expression_matrix)