It seems that the default prediction buffer is sufficient for default 10X segmentation (they seem to generally overestimate cell size), but insufficiently small if membrane markers were used.
1st image:
- Default:
prediction_graph_buffer_ratio=0.05
- Red line: membrane marker-based segmentation
- Turquoise dots: Candidate transcripts for segger segmentation
2nd image:
- Increased Buffer:
prediction_graph_buffer_ratio=0.4
Perhaps we could use the metadata to update the default parameter depending on wether a membrane staining was used. Furthermore, consider adding this to some future type of segger-qc output.
It seems that the default prediction buffer is sufficient for default 10X segmentation (they seem to generally overestimate cell size), but insufficiently small if membrane markers were used.
1st image:
prediction_graph_buffer_ratio=0.052nd image:
prediction_graph_buffer_ratio=0.4Perhaps we could use the metadata to update the default parameter depending on wether a membrane staining was used. Furthermore, consider adding this to some future type of segger-qc output.