search

cole-trapnell-lab / monocle3

Other
340 stars 102 forks source link

interpretation of trajectory and batch effect #440

Open Tushar-87 opened 4 years ago

Tushar-87 commented 4 years ago

Dear All,

I am struggling to understand the trajectory. I read that black circles are nodes and the grey ones are their destinations, but I could not find the meaning of numbers in those circles.

I preprocessed adult mouse endothelial cells, aligned the 2 samples to remove the batch effect and then clustered the cells (exact protocol mentioned at Monocle3 website). I get three clusters denoting artery, vein and capillary endothelium.

When I split the cells into underlying samples (wt and ko), I see the same clusters. However, the cell arrangement in the clusters differs between the samples.

I do not see this when I cluster the cells using Seurat (after integration step). I am wondering whether Seurat performs over correction of the batch effect or Monocle3 underperforms the batch correction.

Please find the attached graphs. Kindly help me to understand these results.

Best wishes, Tushar Monocle3 endothelial cells.pdf

hpliner commented 4 years ago

The numbers in the circles are merely for labeling/reference purposes and don't have any special meaning. I'm not sure I understand your second question, but any time you recluster on a different set of cells you will get somewhat different clustering. Not sure if that answers your question

brgew commented 3 years ago

Is this resolved?

Tushar-87 commented 3 years ago

Is this resolved?

The first issue-This has been answered by @hpliner. So basically one can ignore the numbers in the graphs.

The second issue: Why is there different arrangement of cells in the wt and ko samples? I guess one can interpret that the clustering pattern is stochastic. However, we see differences in the arrangement of cells between wt and ko (slide 4). We have already corrected for batch effect. Then the question remains: is the visible difference biological?

The reason I am worried about the apparent difference in the cluster/cell arrangements between 2 groups is that we do not see this when the samples are integrated and clustered in Seurat pipeline.

andynkili commented 2 years ago

@Tushar-87

If numbers can be ignored, how can one know the direction of the trajectory? I thought those number labels were indicating the direction in the graph.

Best, Andy

jon-xu commented 2 years ago

You can plot the pseudotime(cds) against the clusters(cds), which can show you which cluster is earlier which is later. But I also don't get it why we need the numbers if it's not the time sequence...