# 2023/10/26 SpatialData community meeting [Zoom link](https://embl-org.zoom.us/j/91981143644?pwd=M2NuRjlueVhqbmd4VDZFMktyWjAwdz09) 6PM to 7PM CET If you are an external guest and you plan to attend, please write your name among the attendees and add time for a short introduction or topic to the agenda. Should you wish to attend from a different time zone, please inform us here of your time-zone and topic so we can plan a more convenient slot. *Attendees:* Name, Brian Long (Allen Institute for Brain Science), Luca Marconato, Quentin Blampey, Benjamin Rombaut, Wouter, Nick Fernandez, Giovanni ## Agenda and notes Please add topics of discussion and include a time estimate - (Brian Long) Learning about expectations for `spatialdata-io` stability (5-10min) - External (commercial) vendor support/investment for up-to-date readers? - focus on Xenium and MERSCOPE (Michael). Loading data via spatialdata-io, but some problems with data loading. - (Quentin Blampey) experience with Xenium and MERSCOPE. MERSCOPE laoder works on their part. - (Nick Fernandez) Computational Biologist, formerly at Vizgen. Interested in open source. - (Luca) Focus on stability and bug fixing first, then new features. de.NBI CI/CD system for testing. - limitation is not testing all version changes when companies release updates, would be lot of work. Solution would be for companies to make the PR to add this support. - mentions https://github.com/quentinblampey/spatialdata_xenium_explorer - old HDF5 MERSCOPE format currently note supported. - (Nick) converter is available to update old format to newer Parquet format. (https://vizgen.github.io/vizgen-postprocessing/) - (Giovanni) need for more small open datasets from the companies to check for issues. - (Luca) transformation, table, geometries not in NGFF spec. Internally for spatialdata not a problem. Spatialdata implementers (vitessce, spatialdata in R) can use the test spatialdata datasets. - (Luca) no spec for multiscale geometries. Solution: subsampling (napari-spatialdata) or datashader (spatialdata-plot). - (Luca) for larger datasets need for spatial indexing of GeoPandas (future meeting with GeoPandas developers) - (Nick) There’s a format used in geospatial work called Mapbox tiles that does multi-level representations of points and polygons, but it only works in geospatial coordinates (https://observablehq.com/@cornhundred/deck-gl-mvt-layer?collection=@cornhundred/deck-gl) - (Nick) 3D segmentation. Multipolygon at a single z instead of only polygon (support for multiple cells at a location in a tissue). - (Luca) a lot of support for Multipolygon already. But not for 3D, just 3D labels. - (Nick) using one .parquet file instead of multiple ones and using row groups is more performant for filtering operations. Custom sorting can be used to improve performance ## Resources - link to spatial muon: - very early version with (untested) implementation of trimesh for 3D regions support: https://github.com/ilia-kats/spatialmuon/blob/8d6e9f12cf2ab2ed5a1b231cc7818714cd28020a/spatialmuon/datatypes/array.py - latest implementation, the framework is more tested and stable, but the 3D support has been removed; it appears commented in some parts of the code like here: https://github.com/ilia-kats/spatialmuon/blob/202202_graphs/spatialmuon/datatypes/regions.py ## Actionable items - (Luca) DONE make 10x aware of https://github.com/quentinblampey/spatialdata_xenium_explorer - (Luca) DONE ask 10x and Vigzen for informing on datasets changes, or for providing small datasets - (Luca) DONE share the Mapbox tiles example to the NGFF community, Vitessce dev and geopandas developers - (Luca) DONE discuss regarding spatial indices for speeding up computation with the geopandas devs. Answer: using the Hilbert curve ordering and saving this data will speed up things. - (Luca) DONE prepare roadmap document and share it - (Luca) DONE share spatial muon code to Nick # Shareable invitation to this meeting ## SpatialData Community Meeting: Thursday, Oct 26th. 6PM-7PM CET You are welcome to join the next SpatialData Community Meeting on Thursday, Oct 26th. 6PM-7PM CET. - [zoom](https://embl-org.zoom.us/j/91981143644?pwd=M2NuRjlueVhqbmd4VDZFMktyWjAwdz09) (passcode: 570601) - [agenda](https://hackmd.io/C__F_rwXRz-WgCF1YFuiQg) General information: - [all the future meetings and agendas](https://hackmd.io/enWU826vRai-JYaL7TZaSw?both) - [zulip chat](https://scverse.zulipchat.com/#narrow/stream/315824-spatial/topic/spatialdata.20meetings)