Loading setup…
Configure a synthetic analysis
Choose Illumina or Oxford Nanopore, then Fresh or FFPE. Select input files, assign samples and choose analysis settings. Save the configuration, start or stop analysis, follow its progress, and open the results here.
Call variants from existing BAMs →Continue an existing analysis
Choose your saved analysis output project containing config/run.yml and results, including after moving it. Stop the project's current run first.
Leave blank to use the saved project's image. Correct an unavailable tag here before loading.
Saved samples and analysis settings are retained. Analysis starts only after review and Run. Compatible completed stages are checked before reuse.
Assign samples to a CpG classifier
Check this classifier group
Advanced · original inputs, tools and retained settings
Sample names, original FASTQ groups and reference come from the ONT project. Review or correct the original modified-base input and installed resource paths here. Saving performs the full checks; no analysis starts automatically.
Choose your sequencing platform
Choose sample preservation
Choose how these specimens were preserved.
Use one preservation type for the study specimens in each project. Both platforms support either choice.
Find your FASTQ folder
Browse the synthetic examples, or use Browse computer to preview sequencing filenames in a folder you choose. Run OncoTracer locally to analyze your files.
Select a run to fill its matching folders, or browse each path below. Review the paths before saving.
Not checked — discover FASTQs or browse to your POD5 folder.
Not checked — discover FASTQs or browse to your modified-base BAMs.
Link completed POD5 or modified-base BAMs from this run for methylation. Use files MinKNOW has finished writing. FASTQs define each barcode's read IDs; only matching reads enter its methylation profile. A ligation FASTQ folder defines one sample. Choose which signal source to use in analysis settings.
Samples in the selected folder
| Detected sample / barcode | FASTQ files |
|---|
Assign samples in the columns below. Illumina read pairs stay together; ONT batches stay grouped by barcode. Filtering this list does not exclude samples.
Assign and name your samples
Drag detected samples into Normal or Cancer, or use the type dropdown on each card. Drag between columns to correct a type, or back to Unassigned to exclude a sample. Edit the detected names as needed.
Normal samples are analyzed independently for CNA. Matched somatic calling requires explicit pairing in variant settings. A custom tag also needs a study or control role.
Detected · unassigned
Normal
Cancer
Custom tags
Choose analysis settings
Default: 100 kb for QDNAseq.
Normal/control samples keep their labels with either CNA method. Each sample is analyzed independently; these controls do not form a panel of normals.
Choose parallel sample jobs
Each job processes one sample. Choose the number of samples running at once and see the cores and RAM allocated to each. Completed outputs are validated before reuse.
This overrides the suggested fits. The count is limited only by the number of samples. Too many jobs may slow processing or exhaust memory.
Detect resources to compare sample-job combinations. Cores here mean CPU threads available to this process; other analyses share this server.
View all study contexts and when to use them (opens in a new tab)
Uses the included evidence catalog. No publication searches or language-model downloads.
ONT methylation with Modkit
Dorado calls modified bases from POD5, or reuses MM/ML calls in your BAMs. Modkit extracts CpG methylation. Select Sturgeon for CNS tumours or MARLIN for leukemias.
Assign samples to a CpG classifier
Choose from the specimen diagnosis: MARLIN for leukemia research or Sturgeon for CNS-tumor research. Drag cards or use their dropdowns. These assignments are separate from Cancer/Normal roles. Leave unknown or unsuitable samples in Unknown / do not classify. Only the classifier group selected below is saved and run; the other group is not queued. Use a separate project for each classifier, keeping the completed ONT project unchanged.
Unknown / do not classify
MARLIN · leukemia
Sturgeon · CNS
Modkit and Sturgeon use CPU. GPU can accelerate POD5 basecalling and MARLIN inference. FASTQs alone do not contain methylation calls.
Or select installed tools and model files below. Blank tool fields use your command search path where supported. Models are required here or in the YAML. Setup records checksums automatically.
POD5 basecalling models
Sturgeon · CNS tumours
Use the pretrained general CNS model and matching hg38 probes, or supply your own paths below.
Read the Sturgeon license. Download and use require your acknowledgement. OncoTracer does not grant a license or redistribute the model.
MARLIN · leukemias
Analysis tools and reference
Prepared downloads are reused. About 8.0 GiB for Illumina or 9.7 GiB for ONT, under your project’s reference folder.
Use a local image tag or a registry image reference. The image supplies analysis tools; selected source files, models and annotation databases stay on this computer and are mounted for the run.
Specimen and analysis
Choose the preservation and the analyses to run.
Preservation: Choose in step 2
Choose Fresh or FFPE from your specimen records.
Complete GLIMPSE2 analysis · Sequencing calls, mandatory germline imputation, ClinVar and available ANNOVAR annotation, traits, polygenic scores, ancestry, homozygosity, interpretation reports and an annotated cohort VCF. Available completed work is verified before reuse.
Imputation adds germline evidence. Absence from the reference panel does not establish somatic origin. FFPE damage flags remain visible, including when imputation supports the allele.
Match sample identifiers exactly. Provide name, age or birth year and sex when known; samples with missing metadata are included.
This setting changes trait/GWAS reporting only. Clinical evidence and polygenic-score filters retain their own requirements.
Variant callers
Matched normals for Strelka2 somatic
Select the normal specimen from the same patient for each tumor. Normal/control group membership alone does not establish a match. The matched normal does not inherit the tumor’s FFPE setting.
How caller results are used
Mutect2 and ClairS-TO produce tumor-only candidates. Strelka2 somatic uses your explicitly matched tumor–normal pairs. Strelka2 germline, FreeBayes, bcftools and Clair3 run independently per sample. Group assignments never create matched pairs automatically.
Resources and advanced settings
Caller tools and models
Start with detect_tools. Open paths only to review or change them.
Detect local tools, or install the missing caller environments.
Simulate resource discovery in fictional /demo folders; your computer is not inspected.
Missing tools detected. Choose how to install them.
Parallel sample resources
CPU and RAM budgets are shared across samples. Auto checks CPU affinity, container limits and available RAM before running. Illumina complete variant analysis needs at least 32 GiB per simultaneous sample. Saved explicit limits are retained.
Existing GRCh38 panels are reused. Missing public resources download at run time with resumable transfers and integrity checks.
Leave empty for the bundled workflow, or specify a compatible local workflow folder. Remote repository runs use the top-level Nextflow entry.
Caller environment
Leave blank to use tools already available to OncoTracer.
Strelka2 environment
Autodetect the separate Strelka2 environment. Docker includes this runtime.
Clair3 model
Check this selection against your sequencing records. Fresh/FFPE does not determine the flow cell or basecaller. The matching model is about 75 MB, downloaded once and reused.
ClairS-TO model
Uses the selected model from the ClairS-TO installation or container. Check the flow cell, basecaller version and sampling rate against your sequencing records.
Use an existing ClairS-TO container (optional)
An alternative to the installed executable. Requires local Apptainer or Singularity.
Target regions (optional)
Choose the BED for your assay and reference build. Target regions are not guessed automatically.
Variant filtering
Caller quality filters are applied automatically.
FFPERASE artifact assessment
Existing paths are reused. Public resources are verified and cached in this project's results folder. The FFPERASE runtime still needs to be installed, or use Docker.
Review the FFPERASE license before enabling automatic preparation. The license excludes clinical reporting and requires MSK permission to publish research results. Existing local resources can be supplied below.
FFPERASE source and models
FFPERASE runtime
The standard model assesses whether a variant is present. No YAML file or sample key is needed. Default local false discovery rate: 5%.
Advanced evidence settings
For custom statistical models only. Use the event and sample keys defined inside that YAML.
This is the model's internal sample key, not a patient or barcode identifier.
Annotation
Use your local ANNOVAR installation and matching databases.
Installation and database paths
Requires a matching reference build and an unpacked RefSeq TXT / Mrna FASTA pair. Optional ClinVar uses the same build.
If unavailable, calling results are retained and annotation is reported as unavailable. Installation help appears in resource details.
Save project settings
The selected folder is used directly; no extra project folder is added.
A new project saves the settings above. An existing project loads its saved samples and settings for review and resume.
Saved assignments for this run
Read-only snapshot reviewed before this run started. These are the assignments used by the analysis.
| Analysis sample | Sample type | Analysis role | Detected sample / barcode | FASTQ files |
|---|
Review and run
View the simulated configuration
Simulate Run plays an illustrative progress animation. It does not validate resources, install tools, download references, process reads or produce scientific results.
Choose parallel sample jobs
Each job processes one sample. Choose the number of samples running at once and see the cores and RAM allocated to each. Completed outputs are validated before reuse.
This overrides the suggested fits. The count is limited only by the number of samples. Too many jobs may slow processing or exhaust memory.
Detect resources to compare sample-job combinations. Cores here mean CPU threads available to this process; other analyses share this server.
Run analysis also applies and checks these choices before starting.
Your simulated workflow
These are the settings you explored. No alignment, CNA call, variant, probability, genotype or clinical interpretation was generated.
A real analysis produces caller VCFs, evidence tables, quality-control summaries and reports. Download and run OncoTracer to process your own data.
Install OncoTracer →