Fixes found by running rsemflow on real ENCODE RSEM output
(validation/encode-k562-knockdown/).
annotation ensembl: genes are mapped only when the OrgDb returned a
symbol, Entrez ID, or description. Previously, Ensembl IDs absent from the
OrgDb were also reported as mapped (22,350 of 59,526 genes in the
validation data).enrichment gsea runs serially without a progress bar. On Windows it
previously printed a progress bar and started a worker process. Results are
unchanged.annotation ensembl no longer prints an empty line when loading the OrgDb.First public release.
data import: RSEM *.genes.results import through tximport, with checks
for required columns, one-to-one sample matching, and identical gene tables.
sample_id is always read as text.data summarize: sample- and gene-level expression summaries.annotation ensembl / annotation gtf: gene annotation that keeps the
original RSEM gene ID.expression normalize / pca / correlation: VST, size-factor normalized
counts, or log2(TPM + 1); PCA tables; sample correlation.differential deseq2: metadata-driven DESeq2 with explicit formula, term,
and reference levels; multi-group, numeric, interaction, simple-effect
(--p-at), and LRT analyses. Coefficients are resolved from the design
matrix, and full-rank designs are required.enrichment gsea / gsva: preranked GSEA (fgseaMultilevel, fixed seed)
and GSVA/ssGSEA scores with MSigDB (msigdbr, species-aware collection
aliases) or custom GMT gene sets.effective_length = 0 are handled when building DESeq2
objects.