- fixed edge case bug where infer_type_I_probes fails because there are no type-I-red probes - fixed 16-bit overflow bug reading IDATs with fluorescence > 32766 units created negative values; Reason: pandas version > 1.2.5 can't coerce the uint16 numpy array into float32, but casting it as int64 first resolves the negative fluorescence errors.
1.7.0
- added support for `parquet` file formats (as an alternative to pickle that is readable by other languages) - run_pipeline (beta, M, noob, raw_meth/unmeth, samplesheet_data_frame) as parquet - minor fixes to GEO download processing
1.6.2
- Minor bug fixes
1.6.1
- samplesheet: ensures csv and meta_data pickle match - better error message when multiple samplesheet csvs are detected, and more stringent detection parameters - updated CI/CD with github actions, faster testing, dropped flaky unit tests - updated documentation
1.6.0
- qualityMask: All versions of 1.5.x used a different convention for probes to drop or keep. In version 1.6.0 and above, methylprep reverts to the previous convention used in v1.4.6 and below: For the `quality_mask` column in the processed output CSV and the in-memory SampleDataContainer data_frame, 1.0 means keep the probe (good) and 0.0 means drop. (This reverses a previous convention and fixes bugs where the 0.0 were sometimes NaN.)
1.5.9
- Fixed bug: the SampleDataContainer returned by run_pipeline did not have 1.0 and 0 for quality_mask. Matches the CSV export now.