Question of cutomized trajectory and pseudotime calculation

[lengfei5]

Hi tradeSeq team,

Thank you so much for this fantastic package, which is just exactly what I need.

Because I did not use slingshot or monocle to construct the trajectory and pseudotime, I would like to know how to construct

• pseudotime
• cellWeights
  for evaluateK and fitGAM functions from.

In our data, there is one very clear bifurcation and I have fitted two trajectories separately with principle curves.

I have manually assign the cell weights, e.g. shared the starting cells, I assign 1, 1 to both trajectories; after bifurcation 1 or 0.
This rough assignment should be ok ?

Regarding the pseudotime, I got the pseudo time from each of those two principle curve fitting.
However, I realized that for pseudotime input, I need the pseudotime of each cell for both trajectories.
I tested to project the cell into the trajectory they don't belong to and the result is not very good.

My question is that those inaccurate pseudotime for the trajectory cells don't belong to will affect the algorithm ?

Thanks.

[lengfei5]

Hi there again,

I found a similiar question in #146

The answer was:
"Each cell will need to have a pseudotime value in both lineages. Since each cell is assigned to only one lineage (the weights are respectively zero and one), it does not matter what value we pick so for simplicity, we can just duplicate the pseudotime values. But it would work the same if you had zeros, or any value, instead."