compbiomed / celda

Bayesian Hierarchical Modeling for Clustering Single Cell Genomic Data
MIT License
48 stars 14 forks source link

Count matrix checking/processing #347

Closed joshua-d-campbell closed 6 years ago

joshua-d-campbell commented 6 years ago

1) Checking for all zero rows/columns was removed with last celdaGridSearch refactor, but should be added back in, maybe in processCounts. Error should be thrown so the user can decide how to handle 2) If a matrix of type 'numeric' is used as input, celda_C/G/CG fails at the end. Probably due to the order in which processCounts and compareCountMatrix is called in the beginning. Needs to be consistent so that the checksum is calculated on either the original matrix or processed matrix in all functions. processCounts probably needs to be added to the down-stream functions whenever compareCountMatrix is called.

definitelysean commented 6 years ago

Fixed in #353 @joshua-d-campbell most changes were in split_clusters.R and initialize_clusters.R. Also see above each model function's definition, where I've placed a very thin wrapper.