Notes: Writes per-sample *_fastp.html/json; if multiqc is present, also writes 01-qc/multiqc_report/multiqc_fastp_report.html. (Implementation: automatic MultiQC invocation and output path)
10.3.1 Salmon mode
batch_salmon
--index <DIR> (required): salmon index
--path_fq <DIR> (required): directory of FASTQs (after fastq_qc)
--signature <one or more groups> (required; space- or comma-separated; all uses every group)
Groups: go_bp, go_cc, go_mf, signature_collection, signature_tme, signature_sc, signature_tumor, signature_metabolism, kegg, hallmark, reactome
-o/--output <CSV/TSV> (required) Output: suffixed with _MCPcounter; index label ID; separator inferred from extension.
IPS
-i/--input <matrix> (required), -o/--output <file> (required) No extra flags (the expression matrix yields IPS sub-scores and a total score).
10.9 Clustering / decomposition
tme_cluster
-i/--input <CSV/TSV/TXT> (required): input table for clustering.
Expected shape: first column = sample ID (use --id if not first), remaining columns = features.
-o/--output <CSV/TSV/TXT> (required): output file for clustering results.
--features <spec>: select feature columns by 1-based inclusive range, e.g. 1:22 (intended for CIBERSORT outputs; exclude the sample ID column when counting).
--pattern <regex>: alternatively select features by a regex on column names (e.g. ^CD8|^NK). Tip: use one of --features or --pattern.
--id <str> (default: first column): column name containing sample IDs.
--scale / --no-scale: toggle z-score scaling of features (help text: default = True).
--min_nc <int> (default: 2): minimum number of clusters to try.
--max_nc <int> (default: 6): maximum number of clusters to try.
--max_iter <int> (default: 10): maximum iterations for k-means.
--tol <float> (default: 1e-4): convergence tolerance for centroid updates.
--print_result: print intermediate KL scores and cluster counts.
--input_sep <str> (default: auto): input delimiter (e.g. , or \t); auto-detected if unset.
--output_sep <str> (default: auto): output delimiter; inferred from filename if unset.
nmf
-i/--input <CSV/TSV> (required): matrix to factorize; first column should be sample names (index).
-o/--output <DIR> (required): directory to save results.
--kmin <int> (default: 2): minimum k (inclusive).
--kmax <int> (default: 8): maximum k (inclusive).
--features <spec>: 1-based inclusive selection of feature columns (e.g. 2-10 or 1:5), typically cell-type columns.
--log1p: apply log1p to the input (useful for counts).
--normalize: L1 row normalization (each sample sums to 1).
--shift <float> (default: None): if data contain negatives, add a constant to make all values non-negative.
--random-state <int> (default: 42): random seed for NMF.
--max-iter <int> (default: 1000): NMF max iterations.
--skip_k_2: skip evaluating k = 2 when searching for the best k.