Are there any preferred trajectory/pseudotime analyses that do not require an arbitrary start point?

Question

[nathansuek]

Hello, many of the trajectory inference/pseudotime methods require prior knowledge of a starting point. I feel like this creates a lot of bias, especially when the biological basis for that starting point is not clear and can artificially bias conclusions. Two questions:

1. Are there any community preferred methods for trajectory inference that don't rely/rely less heavily on strong biological priors?
2. If not arbitrary, how do those methods identify starting points?

Thanks!

[Jean-Karim Heriche]

Maybe consider seriation/ordination as an alternative.

[nathansuek]

Can you expand upon what you mean by this?

Answer 1

[ATpoint]

slingshot does not require an explicit start, it will estimate one if absent. Still, trajectory tools are not aware of biology, and integrating your biological expertise to find a meaningful start is in my experience a plus. After all, trajectories are fancy ways of fitting some sort of minimal spanning tree through a bunch of data points, I would really aide this analysis with known and citable biology as much as possible to ensure that the trajectories are meaningful.