deeptools

Use bamCoverage with a normalization method such as RPGC or CPM. For RPGC you must also pass --effectiveGenomeSize, for example 2913022398 for human hg38, and set --numberOfProcessors for parallel processing.

For ChIP-seq coverage use RPGC or CPM, and for treatment-vs-control comparisons use bamCompare with log2 and readCount scaling. For RNA-seq use CPM for fixed bins and RPKM for gene-level signal, and never use --extendReads on RNA-seq data.

Run computeMatrix in reference-point mode with --referencePoint TSS against a bigWig signal file and a genes BED file, then plot the resulting matrix with plotHeatmap. You can add k-means clustering and custom color maps for publication figures.

Yes, alignmentSieve with --ATACshift applies the Tn5 offset correction, equivalent to --shift 4 -5 5 -4, using only properly paired fragments. Apply fragment length filters and check the nucleosome ladder pattern afterward.

RPGC requires the --effectiveGenomeSize parameter, and the command fails without it. Use the value matching your assembly, such as 2913022398 for GRCh38/hg38 or 2652783500 for mouse mm10.

deepTools generates coverage tracks but does not perform differential expression or splicing analysis. The --filterRNAstrand option assumes dUTP-style reverse-stranded libraries, so verify library orientation before interpreting strand-specific tracks.