Changes in version 1.29.1 Reduced package size The source tarball is now under 100 MB (94.8 MB), down from over 100 MB. The following files were removed. All of them are archived, with the full Salmon output for all 24 samples, on Zenodo: https://doi.org/10.5281/zenodo.22982581. The with-inf-reps branch on GitHub is a frozen snapshot of the package that still includes them. - The Gibbs inferential replicates (inst/extdata/quants/*/aux_info/bootstrap/, ~393 MB). tximeta() / tximport() on the files in inst/extdata/quants no longer imports inferential replicates. - aux_info/ambig_info.tsv.gz for each sample. - The full Gencode v29 GTF, inst/extdata/gencode.v29.annotation.gtf.gz. New files and datasets - inst/extdata/gencode.v29.annotation.slim.gtf.gz: a slim Gencode v29 GTF with only gene and transcript lines and 6 attributes (5 MB instead of 38 MB). It has no exon lines, so tximeta::addExons() will not work with it. It has a different filename from the full GTF on purpose: tximeta caches TxDbs by GTF filename, so this avoids clashing with a TxDb built from the full Gencode v29 GTF. Code that called makeLinkedTxome(gtf=...) with the full GTF path should switch to the slim filename. Built by inst/scripts/makeSlimGTF.sh. - data(macro_txp_se): a transcript-level SummarizedExperiment with 20 inferential replicates, for all transcripts of 2000 random expressed genes plus GBP1-GBP7. The counts and infRep assays are rounded to 1 decimal. This is now used by the fishpond swish vignette. - data(macro_tx2gene): a transcript-to-gene table for all Gencode v29 transcripts, for use with tximeta::summarizeToGene(se, skipRanges=TRUE, tx2gene=macro_tx2gene). - Both datasets are built by inst/scripts/makeSumExp.R. The script needs a macrophage install that still has the inferential replicates (<= 1.29.0, or the with-inf-reps branch). Unchanged - The gene-level data(gse) object and the quantification files (quant.sf.gz, etc.) in inst/extdata/quants are unchanged.