{{ report_title }} — {{ basename }}

MRD QC Report — Quality control diagnostics: impact of signature and read filters on ctDNA detection.

Detection Result

{% if detection.detected is true %}

{% elif detection.detected is false %}

{% else %} {% if detection.jeffreys_prior_applied %}

ℹ No background reads observed in synthetic controls — noise rate estimated via Jeffreys prior (p_err = {{ noise_rate_str }}).

{% endif %} {% if detection.n_synthetic_controls < 20 %}

⚠ Only {{ detection.n_synthetic_controls }} synthetic controls — p-value reliability is reduced.

{% endif %}

{% if detection.qc_checks %}

QC Checks

{% for check in detection.qc_checks %}

{% if check.passed %}✅{% else %}⚠️{% endif %} {{ check.label }} {{ check.value_str }} threshold: {{ check.threshold_str }}

{% endfor %}

{% endif %}

Assay Metrics

Signature Size

{{ "{:,}".format(detection.signature_size) }}

filtered loci

Mean Coverage

{{ "%.1f"|format(detection.mean_coverage) }}×

at signature loci

Total Coverage

{{ "{:,.0f}".format(total_coverage_n) }}

reads covering signature

PASS-FILTER RATE

Estimated fraction of true events passing read filters

Total Signal

supporting reads pass-filter rate

Patient vs. Controls

Left plot — Left axis: ctDNA VAF (log scale). Right axis: Signature supporting reads (= VAF × corrected coverage).
★ Patient: matched patient signature.
● Synthetic control signatures: Signatures drawn from db_control.
Blue violin: Fitted Binomial, according to synthetic control error-rates: Binom(N={{ detection.n_effective }}, p_err={{ noise_rate_str }}).
⎯⎯ Detection threshold (min VAF for Binomial p < {{ "%.0f"|format(detection.alpha * 100) }}%).
·–· Sample-specific LOD = {{ lod_str }} — minimum VAF detectable at {{ "%.0f"|format(detection.lod_recall * 100) }}% recall / {{ "%.0f"|format(detection.lod_fpr * 100) }}% FPR for this patient’s signature size and coverage.
◆ Cohort control signatures.

{% if sbs6_vaf_plots or sbs96_plots %}

Signature Profile & VAF

{% for img in sbs6_vaf_plots %}

{% endfor %} {% if sbs96_plots %} {% for img in sbs96_plots %}

COSMIC-format SBS96 profile from all filtered signature variants. 96 trinucleotide channels grouped by substitution type, normalised to pyrimidine strand.

{% endfor %} {% endif %}

{% endif %} {% if intersection_af_img %}

cfDNA Intersection Allele Fraction

Allele-fraction distribution at signature loci with ≥1 supporting cfDNA read. Red: patient signature loci (signal). Blue: control signature loci (noise).

{% endif %} {% if read_length_img %}

Read Length Distribution

Read length distribution of cfDNA reads at signature loci (after read filter). Red: patient signature loci. Blue: control signature loci. Shorter fragment lengths are enriched in ctDNA due to nucleosomal positioning. Synthetic control counts are averaged across synthetic signatures.

{% endif %} {% if intersection_snvq_img %}

cfDNA Intersection SNVQ Distribution

SNVQ score distribution of cfDNA reads at signature loci (after read filter, SNVQ ≥ 40). Red: patient signature loci. Blue: control signature loci. Higher SNVQ indicates higher confidence in the SNV call. Synthetic control counts are averaged across synthetic signatures.

{% endif %} {% if lq_fraction_hist_img %}

LQ-Reads Fraction per Locus

$LQ-reads fraction histogram$

Per-locus fraction of reads failing the read filter (LQ reads) for matched signature loci. Blue: loci kept (LQ fraction ≤ threshold). Red: loci excluded by the noisy-loci filter. Dashed line: threshold{% if thresh_noise_lq_reads is not none %} ({{ thresh_noise_lq_reads }}){% endif %}.

{% endif %} {% if supporting_reads_hist_img %}

Alt-Supporting Reads per Variant Locus

Distribution of alt-supporting cfDNA read counts per signature locus (after read filter). Red: patient loci — signal. Blue: control loci — noise baseline. Loci with zero supporting reads are not shown; their count is annotated in the plot.

{% endif %} {% if applied_filters %}

Applied Filters

{% for name, desc in applied_filters.items() %} {% endfor %}

Filter	Description
{{ name }}	{{ desc }}

{% endif %}

QC Analysis — Unfiltered Signature

Filtered reads + unfiltered signatures: assesses the impact of signature QC filters.

{% if unfilt_sig_comparison %} {% for row in unfilt_sig_comparison %} {% endfor %}

Analysis	Call	p-value	Supporting Reads	ctDNA VAF
{{ row.label }}	{{ row.call }}	{{ row.p_str }}	{{ "{:,}".format(row.supporting_reads) }}	{{ row.vaf_str }}

{% endif %} {% if unfilt_sig_signal_noise_img %}

{% endif %} {% if unfilt_sig_sbs_vaf_img %}

{% endif %} {% if unfilt_sig_intersection_img %}

{% endif %} {% if unfilt_sig_read_length_img %}

Read length distribution (unfiltered sig)

Read length distribution of cfDNA reads at signature loci (filtered reads). Red: patient signature. Blue: control loci. Synthetic control counts are averaged across synthetic signatures.

{% endif %} {% if unfilt_sig_snvq_img %}

SNVQ score distribution at signature loci (filtered reads, SNVQ ≥ 40). Red: patient. Blue: control. Synthetic control counts are averaged across synthetic signatures.

{% endif %}

QC Analysis — No SNVQ Filter

Reads passing all filters except the SNVQ threshold (filt>0 and mapq≥60) at filtered signature loci. Dark bars in the SNVQ histogram (above threshold) should match the primary supporting-reads count.

{% if no_snvq_comparison %} {% for row in no_snvq_comparison %} {% endfor %}

Analysis	Call	p-value	Supporting Reads	ctDNA VAF
{{ row.label }}	{{ row.call }}	{{ row.p_str }}	{{ "{:,}".format(row.supporting_reads) }}	{{ row.vaf_str }}

{% endif %}

{% if unfilt_reads_signal_noise_img %}

{% endif %} {% if unfilt_reads_sbs_vaf_img %}

{% endif %} {% if unfilt_reads_intersection_img %}

{% endif %} {% if unfilt_reads_read_length_img %}

cfDNA Intersection Read Length (no SNVQ filter)

cfDNA Intersection Read Length Distribution at filtered signature loci (filt>0 and mapq≥60, no SNVQ threshold applied). Red: patient signature. Blue: control loci. Synthetic control counts are averaged across synthetic signatures.

{% endif %} {% if unfilt_reads_snvq_img %}

SNVQ distribution at filtered signature loci (filt>0 and mapq≥60, SNVQ ≥ 40 shown). Light bars: reads below the SNVQ threshold. Dark bars: reads above threshold (= supporting reads count). Synthetic control counts are averaged across synthetic signatures.

{% endif %}

{% if noise_filter_comparison %}

QC Analysis — Noisy Loci Filter Sensitivity

Impact of the noisy loci filter (lq_fraction > {{ noise_filter_comparison.thresh_lq }}) on detection. Loci where more than {{ "%.0f"|format(noise_filter_comparison.thresh_lq * 100) }}% of reads fail the read filter low-quality read(s) are excluded from the primary result.

{% for row in [noise_filter_comparison.with_filter, noise_filter_comparison.without_filter] %} {% endfor %}

Analysis	Call	p-value	Supporting Reads	ctDNA VAF
{{ row.label }}	{{ row.call }}	{{ row.p_str }}	{{ "{:,}".format(row.supporting_reads) }}	{{ row.vaf_str }}

{% if noise_filter_comparison.no_noise_signal_img %}

Signal vs Noise — Without noisy loci filter

{% endif %}

{% endif %} {% if multi_read_filter_comparison %}

QC Analysis — Multi-Read Locus Filter Sensitivity

Impact of the multi-read locus filter (Poisson Bonferroni-corrected p < {{ multi_read_filter_comparison.thresh_pvalue }}) on detection. Matched loci whose per-locus HQ supporting-read count is unexpectedly high given the estimated tumour fraction (λ = TF × mean coverage) are removed as likely germline or artifact variants.

{% for row in [multi_read_filter_comparison.with_filter, multi_read_filter_comparison.without_filter] %} {% endfor %}

Analysis	Call	p-value	Supporting Reads	ctDNA VAF
{{ row.label }}	{{ row.call }}	{{ row.p_str }}	{{ "{:,}".format(row.supporting_reads) }}	{{ row.vaf_str }}

{% if multi_read_filter_comparison.pre_filter_signal_img %}

Signal vs Noise — Without multi-read filter

{% endif %}

{% endif %} {% if control_profiles %}

Control Signature Profiles

{% for profile in control_profiles %}

{% endfor %}

{% endif %} {% if fragment_length_img %}

cfDNA Fragment Length Distributions

{% endif %}

For Research Use Only. This report is intended solely for research purposes and has not been validated for clinical diagnostic use. It does not constitute a clinically validated test and should not be used to guide clinical decisions without independent clinical validation. Results may not be accurate for all sample types or conditions.