Changed
- Updated the logic for
detect_outliersin phenotype transforms to actually reflect the function
name (before it was returning true for inliers...). - Updated
quantizeanddequantizeto minimize data copying as much as possible. - Updated
LDMatrix.load_rows()method to minimize data copying. - Fixed bug in
LDMatrix.n_neighborsimplementation. - Updated
daskversion inrequirements.txtto avoid installingdask-expr.
Added
- Added
get_peak_memory_usagetosystem_utilsto inspect peak memory usage of a process. - Placeholder method to perform QC on
SumstatsTableobjects (needs to be implemented still). - New attached dataset for long-range LD regions.
- New method in SumstatsTable to impute rsID (if missing).
- Preliminary support for matching with CHR+POS in SumstatsTable (still needs more work).
- LDMatrix updates:
- New method to filter long-range LD regions.
- New method to prune LD matrix.
- New algorithm for symmetrizing upper triangular and block diagonal LD matrices.
- Much faster and more memory efficient than using
scipy. - New
LDMatrixclass has efficient data loading in.load_datamethod. - We still retain
load_rowsbecause it is useful for loading a subset of rows.
- Much faster and more memory efficient than using