Mike Lin issues

Results 78 issues of


Mike Lin

tabix: fast mode to get variant cells only

The middle-out checkpoint procedure can be skipped if we’re ok with sticking ./. in all non-variant cells

enhancement

embeddings vector indexer

Census cell similarity search: experimental Python API for searching given AnnData

Having built the vector indexes on Census embeddings (#694 chanzuckerberg/cellxgene-census#1113), develop the Python API inside `cellxgene_census.experimental` to input an AnnData and identify the most-similar Census cells. This necessitates running a...

Census cell similarity search: finalize how to deploy TileDB vector indexes

The Census cell similarity search is backed by TileDB-Vector-Search indexes of the embeddings. These indexes are themselves TileDB arrays to store on S3. Finalize details of where they should be...

Census cell similarity search: experimental Python API for metadata prediction

Building on chanzuckerberg/cellxgene-census#1114, add another API method that predicts metadata attributes of the query cells based on the similar cells. The first pass can simply be a plurality vote of...

Census cell similarity search: set TileDB-Vector-Search indexing parameters and characterize accuracy+perf

The TileDB-Vector-Search indexing procedure has some settings we can tune. Consult with TileDB on a suitable parameter set for the Census cell embeddings, build indexes using chanzuckerberg/cellxgene-census#1112, and do some...

Census cell similarity search: pipeline to build TileDB Vector Indexes of cell embeddings

Develop a productionizable pipeline to build the indexes for [TileDB-Vector-Search](https://github.com/TileDB-Inc/TileDB-Vector-Search) from the stored Census embeddings (starting with scVI but also UCE, Geneformer, etc.). This consists of some Python code to...