{{ report_title_prefix }} — {{ sample

{{ sample_name }} · {{ assay_name }} · {{ assembly_declared }} · VNtyper {{ pipeline_version }} · Run {{ run_date }} · {{ research_use_statement }}

{# The document says what produced it, beside what it is of. "VNtyper 2.0.21" was in a provenance panel and is now at the foot of the document with the rest of the run's provenance - which is the right place for the run's *settings* and the wrong one for the name of the tool, because a reader who has been handed this file, or found it in a folder years later, needs to know what made it before they need to know its schema version. It is a byline, not a second heading: smaller than the sample name, in the muted ink, on its own line. The version is monospaced because it is an exact string somebody will quote in a methods section. #} {# The gene symbol in italics, as HGNC has it. `` and not ``: this is a typographic convention for a technical term, not emphasis, which is exactly the distinction the element exists for. The symbol and the rest of the title are two autoescaped values rather than one string with markup in it. #}
{{ report_gene_symbol }} {{ report_title_description }} — {{ sample_name }}

VNtyper {{ pipeline_version }}
{# THE MASTHEAD. Scan order is who this is about, what state the pipeline computed, then the sentence the configuration wrote for that state. `data-state` drives emphasis and nothing else. None of the three values it takes - `finding`, `no-finding`, `indeterminate` - is ever rendered as text: they are colour selectors, not a verdict vocabulary, and `tests/unit/test_generate_report.py` asserts that no report prints one. Every word here is either the pipeline's own or the configuration's. #}
{# One `
` per label/value pair, so a pair is one flex item and can never be split across a line break. A value the run did not record is marked as absent rather than drawn like a fact. #}

Sample

{{ sample_name }}

Assay

{{ assay_name }}

Assembly
{{- assembly_declared }}

Region
{{- region_resolved }}

{# The state, in words. The three algorithm chips are built in `screening_summary.state_chips` - a chip is the most compressed thing this report says about a stage, so it is the one most easily misread as a verdict, and choosing its wording is presentation logic that belongs in the pure module (AGENTS.md trap 11). The last two are here because they read booleans the computed state carries: coverage QC has a displayed verdict *and* a screening axis, and they are not the same fact - a run with no coverage step at all passes the axis while its verdict says it was never evaluated. The rule chip appears only when the configuration had no message for the computed state, which is what tells the reader that the sentence below is a fallback rather than the one written for this state. #} {# THE VERDICT AND THE STATE, SIDE BY SIDE. Prose wants a measure and a masthead is 1,132px wide, so a column of sentences capped at 44 and 68 characters left half the panel empty while the chips sat in a row above it using a third of theirs. Neither is a layout: they are two stacks that happened to be written in that order. The sentence keeps its measure - that is what makes it readable - and the chips take the width beside it, where a reader scanning for state finds them on one edge instead of reading across four boxes. Below 60rem the two become one column again and the original order returns. #}

{# The configured message, as the ordered parts it was authored in. Every part is a separate autoescaped element: it used to be one `summary_text` interpolation marked safe, and the only reason it was marked safe was the line-break separators inside it - the whole sentence was exempt from escaping to get a line break. Nothing is reworded here; the split lives in `report_config.json` beside the verbatim message, and `screening_summary.render_segments` pins that the two still agree. #}
{{ screening_segments[0] }}
{% for segment in screening_segments[1:] %}
{{ segment }}
{% endfor %} {% if coverage_not_measured_note %}
{{ coverage_not_measured_note }}
{% endif %} {# WHAT THE RUN NAMED. The second question anybody opens this report with, beside the first answer instead of in column eleven of a nineteen-column table, 1,600px down the page and 700px off its right edge. It is a repetition of rows that stay below it, never a replacement for them: the tables remain the record. Where the rows carry more than one distinct name, `variant_identity` returns every one of them and qualifies none - choosing which name to show at the top would be the report deciding which variant the reader gets, which is the defect this issue exists to remove. #} {% if variant_identity %}

{{ variant_identity.name }} {%- for other in variant_identity.other_names %} {{ other }} {%- endfor %} {% if variant_identity.tier %} {# The letter and the words for it, together and always. The letter is what the tables print and what a consumer matches on; the words are what make it mean anything to somebody meeting it for the first time. The full definition is in the reading key under the tables and in this element's `title`, so the hover is a shortcut to something already printed rather than the only route. #} {# The badge carries both call-specific sentences as its hover text: they explain something already stated, and four lines of prose under a two-character badge is the badge losing an argument with its own footnote. Both are printed by the rule below, because the archived PDF is what gets filed and a reader with no pointer has no hover - a hover here is a shortcut, never the only route. #} Tier {{ variant_identity.tier }} {{ variant_identity.tier_label }} {% endif %}

{% if variant_identity.ambiguity %}

Ambiguity interval

{{ variant_identity.ambiguity }}

{% endif %} {% if variant_identity.repeat_form %}

Repeat form

{{ variant_identity.repeat_form }}

{% endif %} {% if variant_identity.kestrel_name %}

Kestrel name

{{ variant_identity.kestrel_name }}

{% endif %} {% if variant_identity.advntr_name %}

adVNTR name

{{ variant_identity.advntr_name }}

{% endif %} {% if variant_identity.flags %}

Flags

{% for flag in variant_identity.flags %}
{{ flag.token }}
{% endfor %}

{% endif %}
{# Which tier-A condition this call actually missed. "At least one tier-A condition is unmet" is a definition, not an answer: the report printed the letter, printed the flags, and left the reader to know that `motif-context-diverges` is one of the five names `reconcile` tests before it promotes a call. `tier_reason` states the connection, and says nothing at all when the blocker left no flag behind rather than guessing at one. #} {% if variant_identity.tier_reason %}
{{ variant_identity.tier_reason }}
{% endif %} {% if variant_identity.note %}
{{ variant_identity.note }}
{% endif %}
{% endif %}

{% for chip in state_chips %}
{{ chip.label }} {{ chip.value }}
{% endfor %}
Coverage QC {{ coverage_qc_text }}
{% if not screening_state.matched_rule %}
Screening rule Not configured
{% endif %} {# The two depth figures, beside the state they qualify. "How confident should I be" is the third question this report is opened with, and answering it meant holding a depth score from the table and two coverage figures from a section 250px below it. They are numbers, not states, so they carry no tone - a chip's colour means a computed state here and a depth is not one. The caveat on each is what the coverage section below still exists to state, and the reason the two figures are labelled differently: the window mean is not comparable between assemblies and the flank depth is. #}
Mean coverage {{ mean_vntr_coverage }}

Flank depth {{ vntr_flank_mean_depth }}

{# The provenance line: the raw state the message above was selected by, spelled out in existing pipeline vocabulary so I2 is visible to a reader, not only to a test. Each algorithm's word comes from `screening_summary.algorithm_state_text`, because the computed result is only a result when the stage produced one: an absent or unreadable stage reduces to the block's `default`, so printing `kestrel_result` here read "Kestrel: negative" for a stage that never called anything. #} {# THE NOTICES, AS ONE BLOCK. Two things this report must say that are neither a result nor evidence: what the artefact may be used for, and that the alignment under every number in it could not be identified. They were a grey box inside the masthead and a red box outside it, in two type sizes, with a gap between them - two boxes, two treatments and about 150px for four lines of text. One block, one treatment, one row each, distinguished by the marker and the accent rather than by having a frame of its own. Both are always readable; neither shouts over the result above them. The warning is *first* and is the one that carries an accent: an alignment the pipeline cannot vouch for changes how every number below it should be read, which is a stronger claim than the standing use restriction. It used to be the last line of the fourth provenance panel, 1,000px down the page, where a reader who did not see the red was told nothing. `role="alert"` puts it in front of a screen-reader user and "Warning:" carries the same meaning in words, so neither depends on the colour. #} {# Each notice's text is one element, so the marker and the whole sentence are two grid items rather than the label and its text landing on separate rows. #}
{% if header_warning %}
Warning: {{ header_warning }}
{% endif %}
{{ research_use_statement }}.

{# THE EVIDENCE, FIRST. Until now the first variant table began at y=1580 of a 3142px document - 50% of the way down, 1.6 viewport-scrolls, behind 536px of provenance panels and a coverage section. On a 390px phone it began at 60%. It is what the reader came for, so it is the first thing under the answer. Coverage, cross-match and the alignment browser follow it: each of them qualifies these rows, and a qualifier is read after the thing it qualifies. #}
Kestrel Identified Variants
{# Four states, and they are four different facts (#242). A table of calls; a run that genotyped and called nothing; a run whose Kestrel step is recorded but produced no readable result (#212); and a summary with no Kestrel step in it at all, which `vntyper report` can legitimately be handed. The unreadable case is checked before the negative and says in words that it is not one. A stage that failed has established nothing about the sample, so a report that renders it as "no variant detected" states a negative the run never reached - which is the defect this whole issue exists to remove, in the one place a reader would never think to doubt it. `escaped_table_html` returns "" for an empty frame, and that empty string is what this branches on - `to_html` on one produces a headerless, bodyless table that renders as a stray empty box under the heading. The middle state was the commonest report in any cohort and it looked like a crashed pipeline: `output_empty_result` writes one row of the literal string "None" so that `kestrel_result.tsv` has a body, and the report tabulated it as `None None None None None None None None NaN Negative` with a count line above it calling it one Kestrel row. The count line is counted in Python from the frame that was rendered, so it cannot disagree with the table beneath it. DataTables' own "Showing 1 to 3 of 3 entries" footer is switched off: it exists only when the CDNs resolve, and it counted the rows left *after* the filter had removed the others. #} {# The switch that emphasises flagged values, beside the tables it emphasises. It used to sit in a strip of controls between the provenance panels and the coverage heading, 450px above the table it acts on and above a second switch with an unrelated scope - a control nobody looking at a table would find, and two controls grouped by nothing but having been written next to each other. It changes emphasis and nothing else. This switch used to drive a DataTables row-visibility predicate that removed every flagged row from the DOM until the reader ticked it: a single-patient report that narrates a flagged pathogenic variant and then renders it invisible is issue #242, and there is no code path in this file that can hide a results row. (The cohort report keeps its filter; hiding flagged rows is defensible for triage across samples.) Both switches were `display: none` inputs with a Bootstrap `` drawing the control from its pseudo-elements, so neither could be reached by keyboard and neither was in the accessibility tree. The input is now the control. The `for` attribute is kept even though the label wraps it: an explicit association is what a screen reader reads the name from, and it does not depend on the nesting surviving a future edit. `.table-container` is a scroll box holding a third of the table off-screen, and it was `tabindex: null` with no role and no name - so the columns past the right edge were reachable with a pointer and, for a keyboard reader, only as a side effect of focus scrolling onto a sort button. A named, focusable region is what the ARIA practice for a scrollable region asks for, and the arrow keys drive it once it has focus. #} {% if kestrel_highlight %}

{{ kestrel_row_summary }}
Highlight flagged values

{{ kestrel_highlight | safe }}
{{ kestrel_folded_record | safe }} {% elif kestrel_result_unreadable %}
Kestrel genotyping ran, but its result file is missing or could not be read. No Kestrel result is shown, and this is not a negative.
{% elif kestrel_step_recorded %} {# #266. `detected` and `called` are not synonyms here, and this branch is the one place the difference is visible. When a candidate was seen and suppressed, "No variant detected" is simply false -- one *was* detected, it was not called -- and the line below it would then contradict the line above it. The eight new screening messages make the same distinction, and this keeps the Kestrel section saying the same thing they do. The plain-negative wording is untouched: nothing was detected there, and that sentence is shipped copy. #} {% if kestrel_subthreshold_note %}
No variant called by Kestrel in this sample.
{# One line, autoescaped. It says a candidate was seen and suppressed; it is not a call, it carries no statistics beyond its depth score, and nothing downstream reads it. The state it produces, `negative_subthreshold`, is declared a non-finding in `report_config.json`, so the masthead chip and the screening box are not styled as findings either. `generate_report` withholds this whenever the loaded configuration cannot describe that state, so a deployment that cannot explain the line never prints it. #}
{{ kestrel_subthreshold_note }}
{% else %}
No variant detected by Kestrel in this sample.
{% endif %} {% else %}
Kestrel genotyping was not performed for this sample.
{% endif %}
adVNTR Identified Variants
{# The same four states the Kestrel section above distinguishes, and for the same reason. This section had two branches for them: a table, and one disjunction - "adVNTR genotyping was not performed or no adVNTR results are available" - which covered a failed stage by being vague about it. `advntr_available` now means "produced a result", so an unreadable stage no longer falls into the negative branch and says in words that it is not one. #} {% if advntr_available and advntr_highlight %} {# The same toolbar shape as the Kestrel table above, so the column control has the same home in both and a reader learns it once. The flagged-value switch is not repeated: one switch styles both tables. #}
{% if advntr_row_summary %}
{{ advntr_row_summary }}
{% endif %}

{{ advntr_highlight | safe }}
{{ advntr_folded_record | safe }} {% elif advntr_result_unreadable %}
adVNTR genotyping ran, but its result file is missing or could not be read. No adVNTR result is shown, and this is not a negative.
{% elif advntr_step_recorded %}
No pathogenic variants identified by adVNTR.
{% else %}
adVNTR genotyping was not performed for this sample.
{% endif %}
VNTR Region Coverage
{# The switch that swaps the two coverage views, inside the section it swaps. It used to be paired with the flagged-value switch in one strip 100px above this heading, so a reader looking at the coverage figures had the control that reveals the other eight of them out of sight behind them. #}
Show detailed coverage

Mean Coverage: {{ mean_vntr_coverage }} {{ mean_vntr_coverage_icon|safe }}
Over {{ vntr_region_length }} {%- if vntr_region_length != not_calculated %} bp{% endif %}, as this assembly represents the region. Not comparable with another assembly's mean — the builds contain different amounts of the repeat array.

Flank depth: {{ vntr_flank_mean_depth }}
Mean depth over {{ vntr_flank_bases }} {%- if vntr_flank_bases != not_calculated %} bp{% endif %} of unique sequence either side of the repeat array. This is the figure to compare between assemblies, because unique sequence measures the sample rather than the reference.

Coverage Metric Value

Mean Coverage {{ mean_vntr_coverage }} {{ mean_vntr_coverage_icon|safe }}

Median Coverage {{ median_vntr_coverage }}

Standard Deviation {{ stdev_vntr_coverage }}

Minimum Coverage {{ min_vntr_coverage }}

Maximum Coverage {{ max_vntr_coverage }}

VNTR Region Length {{ vntr_region_length }}{% if vntr_region_length != not_calculated %} bp{% endif %}

Uncovered Bases {{ vntr_uncovered_bases }}{% if vntr_uncovered_bases != not_calculated %} bp{% endif %}

Percentage Uncovered {{ percent_vntr_uncovered }} {%- if percent_vntr_uncovered != not_calculated %}% {{ percent_vntr_uncovered_icon|safe }}{% endif %}

Flank depth (comparable between assemblies) {{ vntr_flank_mean_depth }} — mean over {{ vntr_flank_bases }} {%- if vntr_flank_bases != not_calculated %} bp{% endif %} of unique sequence either side of the repeat array, under counting policy {{ depth_counting_policy }}

VNTR array depth sum {{ vntr_array_depth_sum }} over {{ vntr_array_length }} {%- if vntr_array_length != not_calculated %} bp{% endif %} of array as this assembly represents it

{% if depth_sum_reference_length != not_calculated -%} VNTR array depth per {{ depth_sum_reference_length }} bp {%- else -%} VNTR array depth per reference length {%- endif %} {{ vntr_array_depth_sum_per_unit_length }} — comparable between assemblies only under counting policy {{ depth_counting_policy }}

Coverage QC {{ coverage_qc }}

{% if fastp_available %}
fastp Quality Metrics

Sequencing Setup: {{ sequencing_str }}

Metric Value Status

Duplication Rate (Cutoff: {{ duplication_rate_cutoff }}) {% if duplication_rate_display is not none %} {{ duplication_rate_display }} {% else %} N/A {% endif %} {{ duplication_rate_icon|safe }}

Q20 Rate (Cutoff: {{ q20_rate_cutoff }}) {% if q20_rate_display is not none %} {{ q20_rate_display }} {% else %} N/A {% endif %} {{ q20_icon|safe }}

Q30 Rate (Cutoff: {{ q30_rate_cutoff }}) {% if q30_rate_display is not none %} {{ q30_rate_display }} {% else %} N/A {% endif %} {{ q30_icon|safe }}

Passed Filter Rate (Cutoff: {{ passed_filter_rate_cutoff }}) {% if passed_filter_rate_display is not none %} {{ passed_filter_rate_display }} {% else %} N/A {% endif %} {{ passed_filter_icon|safe }}

{% endif %} {# The screening summary is the masthead at the top of this document. It used to be repeated here as a `.summary-box` under a heading of its own; the message, the state and the provenance line all moved up, so the reader is told the outcome once, before the evidence rather than after it. #}
IGV Alignment Visualization

{# `aria-labelledby` points at the summary, so the disclosure's group takes its name from the words already on screen and cannot drift from them. Chromium does not derive a group name from `
` on its own, and an unnamed group is one a screen-reader user cannot tell from any other region of the page. #}

Variants

{# **The alignment panel's first paint is authored, not blank.** Expanding the embedded library is asynchronous, so there is a real interval - and, for a browser that cannot expand it at all, a permanent state - in which this frame holds no alignment view. Until #242 that interval rendered as an empty box: a failed load threw before any fallback was reached, and the reader was given a panel-shaped hole with nothing in it. Which sentence belongs here is decided in Python, from facts Python has: whether the run produced an alignment session at all, and which `--report-igv` mode was in effect. Writing it into the markup rather than from the script is what makes it true for a reader with scripting off, and it is why there is no spinner: a spinner is a promise that something is coming, and in two of these four states nothing is. #}
{# `off` is asked first, and the order is the point: it is the one state the *operator* chose, so it is the one the reader is owed a reason for. A run switched off has no session either, and answering "no alignment session was produced" would be true and useless - it would read as something the run failed to do rather than something it was told not to. `sidecar` is asked *after* the session check for the mirror-image reason: with no session there is no `igv_report.html` beside the report, and pointing at a file that was never written is worse than saying nothing. #} {% if igv_mode == "off" %}
Alignment visualisation was switched off for this run. --report-igv off was in effect, so no alignment browser was produced. The variant tables above are complete without it.
{% elif not igv_session_available %}
No alignment visualisation is available for this sample. This run produced no alignment session, so there is nothing for a genome browser to show. The variant tables above are complete without it.
{% elif igv_mode == "sidecar" %}
The alignment browser is a separate file. igv_report.html, written beside this report, is a self-contained alignment browser for this sample. It makes no network request either.
{% elif not igv_payload %}
The alignment browser is not in this file. The embedded library is missing, so the alignment view is not shown. The variant tables above are complete without it.
{% else %}
Alignment view igv.js {{ igv_version }} is embedded in this file, compressed, and expands when the report is opened. It makes no network request.
{% endif %}
{# The alignment view is a canvas, and `clearAlignmentState` removes the authored paragraph beside it on a successful load - so without a name here the reader who cannot see it goes from a described state to an unnamed region. The name is on the element that outlives the swap. #}

Reading key
{# THE READING KEY, as an appendix. Every coded value the results tables print, in words, at the foot of the document. It was above the first table for a while, on the argument that a definition belongs before the term it defines. Wrong, for this reader: they open this report knowing what a tier is, and a block of definitions between the answer and the evidence is the same defect as the provenance panels that used to sit there. It is looked *up*, not read through - so it goes where a reader goes to look something up, and the call-specific version of the same two sentences is on the tier badge as hover text where the name itself is. `Tier B` and `known-variant;motif-context-diverges;position-ambiguous` were printed with nothing anywhere in the artefact saying what either meant - and this is a file that is forwarded, archived and opened offline years later, so "look it up" has no object. `nomenclature_legend` lists only the terms these rows carry: a key naming all three tiers and all ten flags under a table that uses two of them is a wall the reader has to filter themselves. It is printed rather than put behind a hover, for the same reason the token layer gives about the motif sequence: this report is read on paper, and a reader with no pointer has no `:hover` at all. The `title` attributes on the column headings and the tier badge are a shortcut to this block, never the only route to it. #} {% if nomenclature_legend %}

How to read the coded values in the tables above

{% for entry in nomenclature_legend %}
{{ entry.term }}{% if entry.label %}{{ entry.label }}{% endif %}

{{ entry.meaning }}
{% endfor %}

{% endif %} {# RUN PROVENANCE, at the foot of the document. Nothing here is dropped and nothing is behind a disclosure: an archived report that cannot say what produced it is not a record, and every value below is still printed, still labelled and still on paper. What changed is where it is and how much room it takes. These ten facts used to stand between the masthead and the first result in four stacked panels at one value per paragraph, over 536px - 17% of the whole document, and a full viewport of settings and digests in front of a reader who came to see a variant. The BAM-header warning is the one line that did not come down here with the rest: it is directly under the masthead, because an alignment the pipeline cannot vouch for changes how every number in this report should be read. #}
Run provenance
{# The raw state the masthead's message was selected by, in the pipeline's own vocabulary. It is provenance, so it is filed with the provenance: on screen it was three lines under the chips saying `High_Precision` where the chip beside it already said "High precision", `positive` where the chip said "Positive" and `PASS` where the chip said `PASS` - the same four facts, twice, one in words and one in tokens. It is not dropped, because the two are not quite the same claim: the chips say what the state *means* and this says what it literally *was*, which is what a reader checking why a configured message appeared needs (#242 I2). Nothing here is hidden - it is one section lower, with the schema version and the library digest it belongs beside. #}
Kestrel: {{ screening_state.kestrel_state_text }} · adVNTR: {{ screening_state.advntr_state_text }} · Coverage QC: {{ coverage_qc }}

Input Files: {{ input_files_text }}

VNtyper Version: {{ pipeline_version }}

Pipeline run started: {{ run_date }}

This report rendered: {{ report_date }}

Reference assembly requested: {{ assembly_declared }}

Assembly detected from BAM header: {{ assembly_detected }} {%- if assembly_contig %} (Contig: {{ assembly_contig }}){% endif %}

Region analysed: {{ region_resolved }}

Summary schema version: {{ summary_schema_version }}
{# What carries the alignment browser, and which one. The digest is over the *decompressed* library - the exact source this document evaluates - and is printed in full so a reader can check it against upstream rather than merely glance at it. `report_assets.igv_payload` verifies the same digest in Python before the payload is written into the file, which is where a check can actually run: `crypto.subtle` is unavailable on `file://` in several engines, so verifying in the browser would be a check that quietly does not happen on the machines this report is most often read on. #}
Alignment browser: igv.js {{ igv_provenance }}

{% if reference_key_used or reference_path or reference_source_effective %}
{% if reference_key_used %}
Reference key used: {{ reference_key_used }}
{% endif %} {% if reference_path %}
Reference path: {{ reference_path }}
{% endif %} {% if reference_source_effective %}
Effective reference source: {{ reference_source_effective }}
{% endif %}
{% endif %} {# The aligner. Its warning is not here: it is under the masthead, where a fact that changes how every number below it reads belongs. #} {% if alignment_pipeline %}

Alignment pipeline: {{ alignment_pipeline }}

{% endif %}
Pipeline Log
{# "Show/Hide Log" was an instruction to the reader about a control that did not announce itself. `
` announces its own state, so the summary is now the name of the thing rather than the verb. #}

Pipeline log

{{ log_content }}

Coverage Metric	Value
Mean Coverage	{{ mean_vntr_coverage }} {{ mean_vntr_coverage_icon\|safe }}
Median Coverage	{{ median_vntr_coverage }}
Standard Deviation	{{ stdev_vntr_coverage }}
Minimum Coverage	{{ min_vntr_coverage }}
Maximum Coverage	{{ max_vntr_coverage }}
VNTR Region Length	{{ vntr_region_length }}{% if vntr_region_length != not_calculated %} bp{% endif %}
Uncovered Bases	{{ vntr_uncovered_bases }}{% if vntr_uncovered_bases != not_calculated %} bp{% endif %}
Percentage Uncovered	{{ percent_vntr_uncovered }} {%- if percent_vntr_uncovered != not_calculated %}% {{ percent_vntr_uncovered_icon\|safe }}{% endif %}
Flank depth (comparable between assemblies)	{{ vntr_flank_mean_depth }} — mean over {{ vntr_flank_bases }} {%- if vntr_flank_bases != not_calculated %} bp{% endif %} of unique sequence either side of the repeat array, under counting policy {{ depth_counting_policy }}
VNTR array depth sum	{{ vntr_array_depth_sum }} over {{ vntr_array_length }} {%- if vntr_array_length != not_calculated %} bp{% endif %} of array as this assembly represents it
{% if depth_sum_reference_length != not_calculated -%} VNTR array depth per {{ depth_sum_reference_length }} bp {%- else -%} VNTR array depth per reference length {%- endif %}	{{ vntr_array_depth_sum_per_unit_length }} — comparable between assemblies only under counting policy {{ depth_counting_policy }}
Coverage QC	{{ coverage_qc }}

Metric	Value	Status
Duplication Rate (Cutoff: {{ duplication_rate_cutoff }})	{% if duplication_rate_display is not none %} {{ duplication_rate_display }} {% else %} N/A {% endif %}	{{ duplication_rate_icon\|safe }}
Q20 Rate (Cutoff: {{ q20_rate_cutoff }})	{% if q20_rate_display is not none %} {{ q20_rate_display }} {% else %} N/A {% endif %}	{{ q20_icon\|safe }}
Q30 Rate (Cutoff: {{ q30_rate_cutoff }})	{% if q30_rate_display is not none %} {{ q30_rate_display }} {% else %} N/A {% endif %}	{{ q30_icon\|safe }}
Passed Filter Rate (Cutoff: {{ passed_filter_rate_cutoff }})	{% if passed_filter_rate_display is not none %} {{ passed_filter_rate_display }} {% else %} N/A {% endif %}	{{ passed_filter_icon\|safe }}

{{ report_gene_symbol }} {{ report_title_description }} — {{ sample_name }}

Kestrel Identified Variants

adVNTR Identified Variants

VNTR Region Coverage

fastp Quality Metrics

IGV Alignment Visualization

Reading key

Run provenance

Pipeline Log