Reading your results¶

This page explains what each part of a sample page means when you open a library from All tracked samples (use View on that row). The exact charts and numbers depend on which analyses your team enabled and how far the run has progressed. For a screenshot of the tracking table (with anonymised quality-control examples), see All samples in the tour.

Sample page: library header and actions, Run summary tiles, Classification details, Analysis details

Example only: the library ID and metrics in this screenshot are from anonymised quality-control data; your page will reflect your run.

What you’ll see first: Run summary¶

Run summary is a block of tiles at the top—think of it as a run card for the sequencer:

When the run started and how long it ran
Which device and flow cell were used
Kit, basecalling settings, and other technical metadata

Use it to confirm you’re looking at the right run and right patient before you interpret biology below.

Long values (e.g. a full basecall model name) may wrap across the full width so nothing is cut off.

Classification details¶

Classification details shows how different classifiers (for example Sturgeon, NanoDX, PanNanoDX, Random Forest) rank possible tumour classes or methylation-based groups.

How to read it:

Each card gives a main call and a confidence level (often with a coloured bar).
Higher confidence usually means the model is more certain—but all results still need clinical interpretation by your team.
Click a card to jump to the expanded section below with charts (bar charts of top classes, and sometimes confidence over time).
Reference lines on charts may be labelled Medium and High to match your lab’s reporting thresholds.

Not every classifier appears for every run; it depends on your pipeline settings.

Analysis details¶

Analysis details summarises coverage, copy number (CNV), MGMT methylation, and fusion candidates in a row of cards.

How to use it:

Skim the headline on each card (depth, CNV status, MGMT level, fusion counts).
Click a card to scroll to the longer section below with plots and tables for that topic.
Coverage — whether targets are sequenced deeply enough; your lab may use coloured bands or thresholds (e.g. sufficient vs low).
CNV — gains and losses; plots are for visual review—CNV calls are often heuristic.
MGMT — methylation at the MGMT promoter region; interpretation depends on your clinical protocol.
Fusion — candidate gene fusions; follow-up may use tables or genome views.

Detailed results (full sections)¶

Below the Classification details and Analysis details cards, the sample page continues with larger blocks—each tied to the same topic as the card above. Your workflow may hide some blocks entirely.

Classification¶

The heading Classification groups Sturgeon, NanoDX, PanNanoDX, and Random Forest in separate expandable rows (click to open).

Classification section: summary row, expanded Sturgeon with top-classes bar chart and confidence over time

Example only: calls and charts are from anonymised quality-control data.

Inside each tool: a short summary (top call, confidence, probe or feature count where shown), then charts—typically a bar chart of top classes and, where available, confidence over time with Medium / High reference lines.
Use this area when you need more than the dashboard card shows: full class rankings and how stable the call was as the run progressed.
Which tools appear depends on your pipeline configuration.

Coverage¶

The Coverage block starts with Coverage Analysis: an overall quality label plus global estimated coverage, targets estimated coverage, and enrichment (how much reads concentrate on targets vs genome-wide).

Further down you will usually find:

Per chromosome target coverage — a chart comparing on-target vs off-target depth by chromosome.
Coverage over time — cumulative estimated depth (×) as the run advances.
Target coverage over time — mean target depth over time, with outlier highlighting (values beyond about two standard deviations from the mean per gene, as described on screen).

Use this section to judge whether sequencing depth is adequate overall and whether any time window or gene looks anomalously low or high.

Target coverage¶

Target coverage is the gene- and region-level view for your panel (for example rCNS2, AML, PanCan—whatever your run used). You should see the panel name, a note that targets come from the gene panel BED, and often an info expansion with panel and BED file details.

Target coverage: per-chromosome distribution with outliers and searchable targets table

Example only: panel and gene labels are from anonymised quality-control data.

Gene amplifications: this section is a good place to notice suspected amplifications—genes such as MYC or EGFR (when on your panel) may show up as high-coverage outliers on the per-chromosome plot (often labelled on the chart) and/or in the table’s outlier column (e.g. flagged relative to mean ± 2 SD). Treat these as leads to review alongside the Copy number (CNV) section and your lab’s clinical rules; unusually high target depth alone is not necessarily proof of amplification.

Typical contents include:

Per-chromosome and per-gene views (including scatter, box-style, or bar plots of depth by region).
A searchable table of targets with coordinates, coverage (×), and outlier flags.
Coverage distribution by gene and any threshold bands your site uses to label sufficient vs low coverage.

Use it when you need which genes or regions drove the headline coverage number, not just the genome-wide average.

MGMT methylation¶

MGMT methylation focuses on the MGMT promoter on chromosome 10.

MGMT methylation: promoter summary, locus plot, latest results, and per-CpG table

Example only: percentages and status are from anonymised quality-control data.

Promoter summary — status (for example methylated vs unmethylated), average methylation (%), prediction score where shown, and a short note that results come from per-site data.
Locus visualization — a plot of methylation along the promoter region.
Often a table of individual CpG sites with strand-specific coverage.

Interpretation is protocol-specific; treat this as supporting information alongside pathology and other assays.

Copy number (CNV)¶

The UI uses the heading Copy number (CNV) for genome-wide copy-number views.

Default (genome-wide): with Chromosome and Gene set to All, you see the full genome on the CNV scatter (ploidy) and difference (relative) plots, plus sliders to pan and zoom. The CNV events table below lists segments (gain/loss, arms or whole chromosomes, affected genes, confidence).

Copy number (CNV): default genome-wide scatter and difference plots, events summary

Example only: events and gene names are from anonymised quality-control data.

Plot bin: the Plot bin menu (for example Data default, 500 kb, 1 Mb, …) re-bins the points used for plotting. Choosing a larger bin (such as 1 Mb) often smooths noisy regions so gains and losses are easier to see at a glance; a finer bin can show more detail when you need it. Try a few settings while reviewing the same sample.

Copy number (CNV): same profile with plot bin set to 1 Mb for clearer smoothing

Example: 1 Mb plot bin for visual inspection; your run may differ.

Single chromosome: set Chromosome to one chromosome (for example chr8) to fill the plots with that chromosome only—cytobands and position on the X-axis make it easier to relate calls to bands and genes. Turn Breakpoints to Show when you want vertical guides at called breakpoints on the difference track.

Copy number (CNV): chromosome 8 only with breakpoints visible

Example: focused chr8 view; anonymised QC data.

Also on screen:

Genome-wide profile card — genetic sex, bin width, variance, and gained / lost counts (aligned with the dashboard card).
Controls — Colour by chromosome vs up/down, linear or log Y-axis, Plot bin, Breakpoints show/hide when available.
Main plots — CNV scatter (ploidy) and difference (relative deviation) share genomic position; use together with the events table.

CNV here is for rapid visual screening; it is not a replacement for certified copy-number assays or expert review.

Fusion analysis¶

Fusion analysis summarises candidates from reads with supplementary alignments (split mappings suggestive of rearrangements).

Candidate summary — counts for target panel vs genome-wide pairs and groups.
Target panel — tables (and often plots) restricted to fusions involving your assay panel.
Genome-wide — broader fusion calls outside the panel, if configured.

Use the tables to inspect gene pairs, support, and grouping; follow your lab’s rules for confirming interesting events.

Sample details page (extra tools)¶

This is a separate URL from the main sample dashboard: /live_data/<library-id>/details. Open it from More details on the main sample page (when shown) or via the tour → Sample details. The page focuses on disk paths, IGV, identifiers, and tabular SNP / fusion / target-gene views—not the large classification and analysis storyboards on /live_data/<id>.

Sample details heading with library ID; Test ID appears when ROBIN can read it from the sample’s identifier manifest on disk.
Intro line listing IGV, sample identifiers, SNP tables, fusion pairs, and target genes.
View sample identifiers — opens a modal with manifest-derived identifier fields (where configured).
Back to sample — returns to the main sample page (/live_data/<library-id>).

Output location¶

Sample output directory — full server path to this library’s folder under the ROBIN work directory; status shows Directory found or not found (with the expected path if missing).

Analysis center¶

On some installs, an Analysis center / Deployment card shows which ROBIN center or deployment label the browser session is using.

IGV browser¶

Embedded IGV.js genome browser (Genome: hg38): ruler, ideogram, reference sequence, gene annotations, coverage histogram, and read alignments.
target.bam must exist in the sample output folder for ROBIN to build the interactive viewer; otherwise you see IGV requires target.bam and a reminder to run target analysis first. When tracks load, ROBIN uses target-scoped indexed BAMs—typically target.bam, or (when present) files such as sorted_targets_exceeding.bam / sorted_targets_exceeding_rerun.bam under clair3/, or igv_ready.bam under igv/. Only alignments from those target-region BAMs are shown—this is not a whole-genome alignment view.
SNP table, indel table (when present), and fusion pairs table each offer View in IGV and/or row clicks that move the browser to the variant, indel, or fusion breakpoints so you can inspect pileups in the loaded BAM. The Target genes table does the same for each gene interval.

IGV browser: hg38 tracks and alignments at a locus opened from the SNP table (View in IGV)

Example only: coordinates, gene, and BAM file name are from anonymised quality-control data.

SNP analysis¶

Appears when SNP processing has written clair3/snpsift_output_display.json.
Summary text may include total variants and counts of pathogenic variants.
Filters: PASS only (keep rows with FILTER = PASS), Pathogenic only, optional Min QUAL, and Reset to clear filters. A search box filters the visible rows. The footer may show how many variants match (e.g. “Showing n of N”).
The table lists columns such as chromosome, position, REF / ALT, gene, HGVS.p, annotation, annotation impact, ClinVar significance (CLNSIG), FILTER, QUAL, genotype (GT), whether the row is pathogenic, Details (expand full fields), and View in IGV.

SNP analysis: PASS only and Pathogenic only enabled, filtered table and summary counts

Example only: variant shown is from anonymised quality-control data; your counts and rows will differ.

View in IGV centres the embedded IGV browser (above on the page) on that variant for pileup review.
A separate indel table may appear when indel display rows exist, with the same filters; View in IGV there jumps the browser to the indel locus in the same target-scoped BAM.
If SNP analysis has not finished or the JSON is missing, you see a short data not found message instead of the table.

Fusion pairs¶

Built from processed fusion pickles (fusion_candidates_master_processed.pkl for the panel first, otherwise fusion_candidates_all_processed.pkl for genome-wide data).
Table: fusion pair name, both chromosomes, breakpoint coordinates, supporting reads, and View in IGV. Click a row or the action icon to jump the IGV browser to both breakpoint neighbourhoods (with padding) so you can inspect supporting reads in the target BAM (same embedded view as SNPs/indels).
Footer text may show total fusions and total supporting reads. Empty or missing data shows a clear no pairs / run fusion first style message.

Target genes¶

Data comes from target_coverage.csv if present, else bed_coverage_main.csv.
Table: gene name, chromosome, start, end, coverage (×) with colour badges by depth band, optional search, sortable columns, and View in IGV. Click a row to open that gene’s interval in IGV (with padding).
If neither file exists, this block is omitted.

Sections appear only when the underlying files exist and your workflow enables the relevant steps.

MNP-Flex (if your site uses it)¶

MNP-Flex is provided commercially by Heidelberg Epignostix GmbH. You can only use it under an agreement with Epignostix; they supply the account credentials ROBIN needs to call their service.

Some sites show an MNP-Flex results block in the sample page. It may include:

When results were last updated
Classifier name and version
A hierarchical summary (class / family / superfamily) and quality or MGMT side panels

MNP-Flex results: toolbar, classifier line, hierarchical summary, QC, and MGMT

Example of a successful test output: classifier metadata, hierarchy scores, QC status, and MGMT; your run will show its own values.

A toolbar may show whether the integration is idle, busy, or running. If you never see this block, your deployment may not use MNP-Flex, or credentials may not be configured.

Applying your Epignostix credentials in ROBIN: The ROBIN process that runs the workflow and web UI reads environment variables on the server (not in the browser). Set the username and password Epignostix gave you before starting ROBIN, for example:

MNPFLEX_USERNAME — your Epignostix user name
MNPFLEX_PASSWORD — your Epignostix password

If either variable is missing, ROBIN does not show the MNP-Flex block and will not run the integration.

How you set them depends on your setup: typically export them in the shell before robin … - whoever operates the sequencing machine or server should make them persistent and secure (never commit them to git).

Your administrator may also tune optional settings (for example MNPFLEX_BASE_URL, MNPFLEX_WORKFLOW_ID, OAuth MNPFLEX_CLIENT_ID / MNPFLEX_CLIENT_SECRET, MNPFLEX_SCOPE) if Epignostix instructs you to; defaults match the standard Epignostix app integration.

Reports and downloads¶

When analysis is far enough along, you can usually generate a PDF report for the sample. The file name is typically based on the sample ID and is saved under that sample’s output folder.

Your team may also offer CSV or ZIP exports of tabular data—if enabled, follow the on-screen options and wait for notifications to finish before closing the tab.

If a download fails, check the notification area and ask your administrator to confirm disk space and permissions.

Important reminder¶

ROBIN is for research use and support to clinical decision-making. Classification, CNV, and fusion outputs are not a substitute for full pathological and molecular review by qualified staff.