Do seeds have a significant impact on the processing results of pyscenic?

[The-G-Cat]

When I first ran PyScenic, I did not set a seed for it. Since I wanted my results to have better reproducibility, I ran PyScenic again and set a seed for it this time. However, when I observed the results, the number of regulons varied significantly, and there were also differences in the scoring. Some regulons had a reversed relationship in the scoring magnitude between the two runs.If so, then could it be that the data from these two analyses are no longer reliable?

The dataset I used is scRNA-seq data of mouse. However, I want to focus on the changes in CF, so after clustering the cells, I extracted the CF cells for pyscenic analysis. Will this screening have an adverse effect on my results? Should I process all the data and then select the CF part from it?

Also, I'm very curious about how many regulons are typically obtained from the data analysis in mouse. I got around 80 to 99, but among them, only about 30 to 40 have existing research foundations. Is this quantity normal?