Monocle 2.2.0 fixes some problems in Census, adds new documentation, and fixes a few other minor issues. This is the first official release of Monocle to through Bioconductor, in version 3.4.
Monocle’s algorithm for converting relative expression values (e.g. TPMs) into absolute transcript counts, called “Census” has been once again been re-designed. The new version improves robustness across different protocols and experimental designs. Census is available through the relative2abs() function. Expect values from this function to change, but differential expression results to be largely the same or slightly better.
Major expansion of the vignette, including a whole new section on the theory behind Monocle 2’s new trajectory reconstruction algorithm.
Fixed an problem where reduceDimension() would return different results on repeated runs given the same inputs. The problem was actually in DDRTree in two places: kmeans and irlba. We now call irlba with deterministically initialized eigenvectors and kmeans with deterministically selected rows of the input.
Fixed a problem in classifyCells() related to joining factors and levels. This would generate annoying warnings.
Fixed the check for valid sizeFactors prior to differentialGeneTest(). Without this check, differentialGeneTest would report FAIL on all genes because of factors not having enough levels.
This is the second public release of Monocle 2 through GitHub. You can check out a full list of features here.