scsim-ext

We extend the single-cell RNA-sequencing simulation framework, scsim, with additional functionalities of: (1) enhancing RNA dropout effect, and (2) adding ambient RNA to drops.

scsim

Simulate single-cell RNA-seq data using the Splatter statistical framework, which is described here but implemented in python. In addition, simulates doublets and cells with shared gene-expression programs (I.e. activity programs). This was used to benchmark methods for gene expression program inference in single-cell RNA-seq data as described here

Enhancing RNA dropout

We implement the dropout step described in Splatter by fitting a sigmoid curve through genes’ log mean count and their cell fraction with zero reads, where the sigmoid is characterized by shape and midpoint parameters. To enhance dropout, the sigmoid is shifted (decrease its midpoint) and the adjusted dropout probability is computed. This probability is then used to binomially sample the observed counts. See example notebook.

Ambient RNA

We lend CellBender's model of sample RNA contamination (or ambient RNA). That is, we add to the gene expression of each drop a portion of contaminating RNA and sample its counts consequently. See example notebook.

Specifically, the mean expression of gene $g$ and cell $n$ (or here, the cell in drop $n$) is obtained from scsim as $\lambda_{ng}$. For CellBender’s probabilistic model, the probability of having a cell in the drop $y_n$ is set to 1 (because empty drops are not considered), and $\rho_n$ is set to 0, because zero reads are exogenous to the drop as we account only for ambient RNA (pumped into the drop) and not for barcode swapping. Thus, the CellBender model simplifies to: $$C_{ng}\sim NB(d_n^{cell}\chi_{ng} + d_n^{drop}\chi_g^a,\phi)$$

The two models were combined as follows: $\lambda_{ng} = d_n^{cell} \chi_{ng}$ is the mean expression (similar to CellBender, scsim also uses a log-normal distribution of cell size) $\lambda_g^{-}=avg_n \lambda_{ng}=\chi_g^a$ is the mean true count of gene $g$, that is, the ambient contribution of the gene. Given the fraction of ambient RNA, $f^{drop}$, and the library size, $d_{\mu}^{cell}$, and scale, $d_{\sigma}^{cell}$, used for sampling the cell size, $d_{n}^{cell}$, $d_n^{cell}\sim LogNormal(d_{\mu}^{cell} + log(f^{drop}), d_{\sigma}^{cell})$ is sampled (that is, the same variance is employed as used for sampling the cell content and the mean is set to $log(e^{d_{\mu}^{cell}}f^{drop})$. $\phi$ is sampled as defined in CellBender

Thus, we sample: $C_ng \sim NB(\lambda_{ng}+d_n^{drop}\lambda^{-}_g, \phi)$.