{% macro rate(value) %}{% if value is mapping and 'events' in value %}{{value.events}}/{{value.total}}{% if value.estimate is not none %} ({{'%.1f'|format(100*value.estimate)}}%; 95% interval {{'%.1f'|format(100*value.lower)}}–{{'%.1f'|format(100*value.upper)}}%){% else %} — undefined{% endif %}{% elif value is none %}undefined{% else %}{{value}}{% endif %}{% endmacro %}

Development cohort comparison

Exploratory research. {{ result.limitations }} Missing results stay unavailable. Counting convention: {{ result.count_convention }}; no count conversion applied.

{{ result.sample_count }} declared samples. Requested folds: {{ result.folds_requested }}; seed: {{ result.seed }}.

{% if result.length is defined %}

Length estimation

Status: {{ result.length.status }}.

{% if result.length.metrics is defined %}

Target: total diploid repeat count in the declared convention. Predictions are from held-out biological groups. The final model is fitted afterward and has no approval identity.

{% for name,value in result.length.metrics.items() %}{% endfor %}

Measure	Value
{{ name }}	{{ value if value is not none else 'undefined' }}

Predictions with warnings: {{ result.length.predictions_with_warnings }}. Warnings do not exclude or clip predictions.

{% if result.length.prediction_warning_counts %}{% for code,count in result.length.prediction_warning_counts.items() %}{% endfor %}

Warning code	Predictions
`{{ code }}`	{{ count }}

{% endif %}

{% for row in result.length.rows %}{% endfor %}

Sample	Truth	Held-out estimate	Training mean	Fold	Prediction warnings
{{ row.sample_id }}	{{ row.truth }}	{{ row.prediction if row.prediction is not none else 'unavailable' }}	{{ row.training_mean_baseline if row.training_mean_baseline is not none else 'unavailable' }}	{{ row.fold }}	{% for code in row.prediction_warnings %}`{{ code }}`{% if not loop.last %}, {% endif %}{% endfor %}

{% else %}

No paired length truth was supplied.

{% endif %}{% endif %} {% if result.callers is defined %}

Caller comparison

Status: {{ result.callers.status }}. {{ result.callers.evidence_status }}. {{ result.callers.identity_truth_status }}.

One outcome-independent representative per declared biological group. Sensitivity and specificity retain no-calls in their truth denominators; missing files never imply negative calls. No-call changes remain explicit.

{% for name in ['true_positives','false_positives','true_negatives','false_negatives','positive_no_calls','negative_no_calls','no_calls','sensitivity','specificity','false_positive_rate','fpr_one_sided_upper'] %}{% endfor %}

Metric	Native baseline	Held-out selected policy
{{name}}	{{rate(result.callers.baseline[name])}}	{{rate(result.callers.selected_out_of_fold[name]) if result.callers.selected_out_of_fold else 'unavailable'}}

{% if not result.callers.selection_available %}

No alternative policy was selectable. Where folds exist, the held-out column repeats the baseline.

{% endif %}

Composition and individual callers

{{result.callers.composition}}. This is a native-call comparison, separate from the pipeline screening verdict.

{% for caller, values in result.callers.per_caller.items() %}

{% for name in ['true_positives','false_positives','true_negatives','false_negatives','no_calls','sensitivity','specificity'] %}{% endfor %}

Metric	Baseline	Held-out selected
{{name}}	{{rate(values.baseline[name])}}	{{rate(values.selected_out_of_fold[name]) if values.selected_out_of_fold else 'unavailable'}}

{% endfor %}

Paired held-out changes

Selected minus baseline. Intervals describe held-out predictions and do not include repeated model-selection uncertainty.

{% if result.callers.paired_differences %}{% for name in ['sensitivity','specificity','false_positive_rate','no_call_rate'] %}{% set metric=result.callers.paired_differences[name] %}{% endfor %}

Rate	Eligible groups	Difference	Descriptive 95% interval
{{name}}	{{metric.eligible}}	{{metric.delta if metric.delta is not none else 'undefined'}}	{{metric.interval if metric.interval is not none else 'undefined'}}

{% endif %}

Observed operating points

{{result.callers.operating_points_scope}}. {{result.callers.roc_scope}}. Zero observed false positives do not establish zero population risk.

{% for point in result.callers.operating_points %}{% endfor %}

Policy	Comparator	Cutoff	Sensitivity	Specificity	No-calls
{{point.policy_id}}	{{point.comparison if point.comparison is not none else 'not declared'}}	{{point.cutoff if point.cutoff is not none else 'not declared'}}	{{rate(point.metrics.sensitivity)}}	{{rate(point.metrics.specificity)}}	{{point.metrics.no_calls}}

{% endif %}

Comparison digest: {{result.comparison_sha256}}

Development cohort comparison

Length estimation

Caller comparison

Composition and individual callers

{{caller}}

Paired held-out changes

Observed operating points