Repository navigation
TEcounts Sample QC
On by default (tetranscripts.qc.enabled: true), the per-sample TEcount
tables (results/tecount/{sample}.cntTable) are merged into a feature x
sample counts matrix (results/tecount/counts_matrix.tsv) and a sample-QC
view is produced with DESeq2 (nf-core/rnaseq style) — the same machinery as
the chimera screen's QC stage, driven by the shared sample_qc.R script. The
matrix is restricted to TE subfamilies by default (feature_class: TE,
the view this pipeline is built for; gene and all also available), then
transformed (vst/rlog/log2, tetranscripts.qc.pca_transform) and turned
into a PCA scatter plus a sample-to-sample distance heatmap, written as MultiQC
custom-content JSON and rendered interactively inside multiqc_report.html
(points colored by the sample sheet's condition column). The
tetranscripts.qc filters (min_samples_present, min_total_counts,
min_events) apply only to this QC view — the per-sample cntTables are
never reduced. If too few features pass, the plot rule ships empty
custom-content JSON (the report documents the skip) and a log message instead
of failing. Set tetranscripts.qc.enabled: false to skip the matrix and plots
entirely (the summary barplots below still render — they only need the raw
cntTables).
The same per-sample tables also feed two summary barplots
(results/tecount/qc/tecount_assignment_mqc.json and
results/tecount/qc/tecount_te_class_mqc.json, rendered in the report's
custom-content section): per-sample read counts and % split into genes vs TE
subfamilies, and the TE-subfamily reads broken down by repeat class
(LINE/SINE/LTR/DNA/RC, anything else grouped as unknown). Unlike the sample-QC
view these use the raw cntTables (all features), so they are independent of
feature_class and always reflect every read TEcount assigned — and they are
always produced, independent of tetranscripts.qc.enabled.