Similar Top DEGs Across Distinct Gene Knockouts in scTenifoldKnk

[Lattesnow]

Hi Developer,

I ran scTenifoldKnk using approximately 1,200 cells and 8,000 genes. I performed in silico knockouts of multiple genes, including several transcription factors and metabolic enzymes, which are expected to produce distinct downstream transcriptional effects. However, I consistently observe that the top differentially expressed genes (DEGs) are highly similar across different knockouts, regardless of the biological function of the perturbed gene.

Below is the parameter configuration used in my analysis:

res <- scTenifoldKnk(
countMatrix = counts,
gKO = tf,
qc_mtThreshold = 0.1,
qc_minLSize = 1000,
nc_nNet = 10,
nc_nCells = nc_nCells_use,
nc_nComp = 3,
nc_q = 0.9,
td_K = 3,
ma_nDim = 2
)

Given the biological diversity of the targeted genes, I would not expect such a high degree of overlap in the top DEGs. I would appreciate any insight into potential causes (e.g., network construction, variance structure, parameter choices, or data preprocessing) that might lead to this behavior.

Thank you very much for your time and guidance.

[dosorio]

Hi @Lattesnow, I don't see any issue with the way you are running scTenifoldKnk other than verifying that nc_nCells_use is different from ncol(counts). I would also not expect a high degree of overlap among genes with low functional relatedness.