`finaletoolkit end-motifs`¶

1. About¶

Measures frequency of k-mer 5' end motifs.

Nucleases do not cut cfDNA at random positions — the enzymes fragmenting DNA in plasma have sequence preferences, so the bases immediately at a fragment's 5' end are informative about which nuclease produced it. This step reads the reference sequence at every fragment end and tallies the frequency of each 4-mer.

The resulting 256-entry frequency vector is a cfDNA feature in its own right, and it is the direct input to mds, which collapses it to a single motif diversity score.

2. As the pipeline runs it¶

finaletoolkit end-motifs bams/{sample}.sorted.bam <ref2bit> \
    -q <--mapscore> \
    -k 4 \
    -min <--fragment-minimum> \
    -max <--fragment-maximum> \
    -o end_motifs/{sample}_endmotif.tsv \
    -w <threads> \
    -v

2.1 Arguments¶

input_file

The analysis BAM. value: bams/{sample}.sorted.bam

refseq_file

2bit reference sequence. value: the ref2bit file for the selected --genome build

The motif is read from the reference, not from the read's own bases, so sequencing errors at the fragment end do not perturb the tally. This is why the 2bit file must be the same assembly the BAM was aligned to — see Reference contig filter for how that is enforced on the BAM input path.

-k

k-mer length. value: 4

4-mers give 256 possible motifs, which is the convention in the cfDNA end-motif literature and what the motif diversity score is normalized against. Changing k would change the number of categories and therefore the scale of the MDS.

-q, --quality-threshold

Minimum mapping quality. value: --mapscore (default 20)

-min, --min-length / -max, --max-length

Fragment length window. value: --fragment-minimum / --fragment-maximum (defaults 50 / 500)

Both strands are counted

FinaleToolkit's -B, --no-both-strands flag restricts the tally to one strand, with -n selecting the negative strand. The pipeline passes neither, so the default applies and the 5' ends of both strands contribute. Each fragment therefore contributes two motifs — one from each end — which doubles the counts available and avoids any strand-specific bias in the frequency vector.

Watch for the flag-name difference if you compare this to interval-end-motifs: there the same -B is spelled --single-strand.

3. Output¶

end_motifs/{sample}_endmotif.tsv — a two-column TSV of each 4-mer and its frequency, 256 rows:

AAAA    0.0083
AAAC    0.0041
AAAG    0.0052
...

This file is consumed directly by mds.

4. Controlling it¶

What	How
Fragment length window	`--fragment-minimum`, `--fragment-maximum`
Reference sequence	the `ref2bit` entry of `config/genome.json` for the selected build
Threads / memory / walltime	the `end_motifs` entry of `config/cluster.json`

"end_motifs": {
    "threads": 24,
    "mem": "4G",
    "time": "08:00:00",
    "partition": "norm"
}

The -k 4 k-mer length is fixed in the workflow rather than exposed as a run option, because mds downstream assumes a 256-motif vector.

5. Requires¶

ref2bit for the selected genome build. Both bundled builds provide it.

interval-end-motifs — the same measurement per genomic window instead of genome-wide
mds — collapses this frequency vector to a single diversity score

finaletoolkit end-motifs¶