Closed lidanwu closed 2 years ago
@patrickjdanaher I have created a function to preprocess SMI data and the corresponding vigentte to perform resegmentation process on SMI raw data from end-to-end manner. To use this vignette, user would only need to change the file path to pre-built SMI giotto object and provide the corresponding config_loading.
I have applied this to the TAP data SMI003-MGH you mentioned and the post-resegementation data is stored at \\freya\NAS\lwu\testRun\SpatialTest\giotto_test\SMITAP_proj\SMI003_DavidTing_MGH\fastReseg01
Here I plotted the processing speed for each FOV as a function of the file size of final updated transcript data.frame when running on Dante container. (There's ~15% file size reduction after resegmentation due to the removal of unnecessary columns and extracellular transcripts). As expected, bigger transcript data.frame would take longer time to process (~8.33 MB per min)
The next thing is to enable visualization of resegmentation impacts. Please let me know if you have any comments and suggestions.
BTW, David Ross did report an issue of running this package when python is not under "/usr/bin/python3". This is because I use Giotto function for leiden clustering and Delaunay network generation. I'm not sure what the best way to address this issue. Please let me know if you have any suggestion.
@patrickjdanaher this PR is ready for review again. Thanks.
Finally, here is the wrapper that could do automatic baseline estimation, take the file path to multiple per FOV transcript files as input and save per FOV cell segmentation outcomes into disk.
Please find the examples for this
fastReseg_internalRef
function below: