Dear all,
referring to scRNA-seq analysis and Monocle3 trajectory reconstruction in pseudotime (https://cole-trapnell-lab.github.io/monocle3/) :
we have 10 samples, and Monocle3 successfully infer the trajectories, however the number of genes that are Diff Expressed along the pseudotime is huge (ie > 4000 genes; we are working with 5 time points).
i would like to ask please : how would you filter further the list of genes that are Diff Expressed along the Pseudotime (beside p-value, q-value) in order to keep the follow the genes that are expressed in a larger fraction of cells ?
could this filtering be done based on Moran I statistics only (or, alternatively, i shall write additional scripts that count the number of expressing cells of gene 1, gene 2, gene 3, etc ) ?
thank you,
-- bogdan
0 answers
No answers yet.
Log in to answer this question.
-
reasonable parameters for graph_test from monocle3
written by sleepystudent •
-
find DEG of monocle 3 object using tradeSeq
written by synat.keam •
-
Monocle3 error when passing gene_module_df directly to plot_cells()
written by Lin •
-
Understanding learn_graph and ncenter in Monocle3
written by kkbarratt19 •
-
Problem with interpreting output from Monocle3 graph_test()
written by Angelina_G •
-
scATAC-seq
written by Bogdan
-
heatmap of genes pseudotime in monocle3
written by Bogdan
-
Gene expression heatmap following pseudotime analysis
written by Vitis
-
what to do after pseudotime analysis by Monocle?
written by hsu •
-
Monocle Trajectories Reversed - how best to correct?
written by eoindosullivan •
Can you elaborate on how the differential testing is done? Does Monocle has its own framework or is this something like Wilcox? Do you test against a fold change of zero, and did you inspect the DEGs in terms off effect size (e.g. MA-plots). Do most DEGs have tiny effect sizes? Are outliers a prominent issue, like comparing cluster A to B, both 1000 cells, and in A 20 and in B 5 cells express the gene with all As being higher than all Bs? I am personally an advocate of pseudobulks (that obviously requires biological replicates) so you can perform "standard" DE e.g. with edgeR which saves you from the issues of sparse data and strong significances due to outliers with (overall per cluster) tiny effect sizes.
Thank you; we have obtained good results by using pseudo-bulk approaches too.
Regarding Monocle3 and differential expression, you could please take a look on :
https://cole-trapnell-lab.github.io/monocle3/docs/differential/
(we have used graph_test() function in Monocle3).
Hi Bogdan,
I have a similar issue struggling with, I am wondering if you were able to find a way to filter DE genes showing a smooth change across samples.
Best V
Hi :
i 'd think that one way would be to filter based on the number of counts per GENE at the beginning and end of the trajectory;
i have started to look a while ago into TradeSeq (https://github.com/statOmics/tradeSeq) that has good documentation :)
Thanks, I'll look through it.