Simulate single-cell RNA-SEQ data using the Splatter statistical framework but implemented in python with additional features. scsim2 is an extension of scsim but now allowing an arbitrary number of activity programs to be simulated.
Added on top of this is SCsim_DiffScale within scsim2_diff_librarysize.py which allows GEPs to have different abundances of RNA content which corresponds to having cells expressing certain GEPs having higher overall library sizes. For example:
deprob = [.025]*7 + [.05]*2 + [.025]*2
deloc = [1.5]*7 + [2.00]*2 + [1.5]*2
simmod_catalog = SCsim_DiffScale(seed=112)
simmod_catalog.simulate(ncells=20000, ngenes=10000, nidentities=7, activity_pct=[.3]*4,
activity_min=[.1]*4, activity_max=[.7]*4,
activity_identities=[list(range(1,8))]*4,
pct_doublets=.05, diffexploc=deloc, diffexpscale=1, diffexpprob=deprob,
delta_range_libloc=0)
simmod_catalog.spectra.sum(axis=1)
Would cause the 8th and 9th programs to have a larger library size because of more incorporated genes and higher average differential expression for those genes.