--prior-segmentation-confidence=1 but contradict to prior_segmentation

[SanoKyohei]

・Command
baysor run input.csv :cell_name -m 30 --prior-segmentation-confidence=1 --plot
・Environment:
Baysor v0.6.2 (in docker)

→This should not change gene assignment, however it changed a lot.
In "segmentation.csv" file, "prior_segmentation" and "cell" doens`t match at all.

Could you give me a solution for this?

[VPetukhov]

The cell ids can be reordered, and cells can be expanded. The restriction is that for a given prior cell all transcripts, which were assigned to it in the prior segmentation must stay assigned to the same cell in the final segmentation. Is it violated?

[SanoKyohei]

The restriction has been violated.
But it was partially because CosMx output bug.
There is a bug in the CosMx output where the same transcriptome appears twice at the same x, y, z coordinates. Due to this bug, the restriction that all transcripts assigned to a given prior cell in the initial segmentation must remain assigned to the same cell in the final segmentation is not functioning correctly.

[VPetukhov]

I'll look into it! Were they also assigned to the same cell in the prior segmentation?