job-runner

MedBook-JobRunner

Jobs

GSEA: preranked + leading edge (RunGSEA)

This job runs GSEAPreranked and then GSEA's leading edge analysis to generate a heat map.

Arguments

GSEA requires a gene set and a gene set group as arguments.

Optionally, the user can run the tool with a gene set created in another tool, such as the outlier analysis or Limma. These links/buttons would be found on the origin tool output page and not in the GSEA tool creation UI.

Optionally, the user can provide a phenotype (two sample groups). This will produce a heat map of samples vs. enriched gene sets.

• gene_set_id, gene_set_name: id and name of the gene set
• gene_set_sort_field: the field on which to sort the gene set. If this is not provided, the organic sort is used. (The order with which the gene set was loaded, stored in gene_labels.)
• gene_set_group_ids, gene_set_group_names: ids and names of the gene set groups. If multiple are provided, combine them into a single .gmt.
• gene_set_associated_object: provided if the job is running with a gene set created by another job. This is used to link to the gene set's origin.
• phenotype: optionally provided, produces a heat map of samples vs. pathways. NOTE: not currently implemented
  • sample_group_a_id: id for sample group A
  • sample_group_b_id: id for sample group B
  • sample_group_a_name: name for sample group A
  • sample_group_b_name: name for sample group B
• set_max, set_min, plot_top_x, nperm, metric, scoring_scheme: arguments provided to GseaPreranked

See here for GSEA tool documentation.

Output

GSEA's preranked tool creates a folder of HTML and data files which are linked together via relative hyperlinks. All of the files in this folder are saved as blobs. Leading Edge creates a heat map (png), which is also saved as a blob.

All of these files are stored as blobs with the associated_object being the job object. The GESA HTML folder entry-point is named index.html, and the heat map is named leading_edge_heat_map_clustered.png.

Paired Tumor Analysis (RunPairedAnalysis)

This job computes a signature of differential gene expression between baseline and progression samples from the same patient.

Arguments

• data_set_id, data_set_name: the data set
• primary_sample_labels: sample labels of all the primary samples
• progression_sample_labels: sample labels of all the progression samples

Output

A gene set linked to the job via associated object security with the columns "Genes" and "Differential score". The differential score column is computed by taking the difference between the average values of the progression samples and that of the primary samples.

Limma (RunLimma)

Limma (LInear Models for MicroArray data) runs the R bioconductor package limma. See here for more details.

Arguments

• value_type: value type of the data
• experimental_sample_group_id, experimental_sample_group_name, experimental_sample_group_version: experimental sample group
• reference_sample_group_id, reference_sample_group_name, reference_sample_group_version: reference sample group
• top_genes_count: number of genes to include in the signature

Output

The gene set created by limma is attached to the job via associated object security.

The two PDFs created by limma are stored as blobs, which are attached to the job. The blobs are named Rplots.pdf and mds.pdf.

Single Sample Top Genes (RunSingleSampleTopGenes)

This job computes a signature with the top N or top N percent of genes in a single sample.

Arguments

• data_set_id, data_set_name: data set of the sample
• sample_label: sample label
• percent_or_count: either "percent" or "count". This specifies whether the job will compute the top N genes or the top N percent of genes.
• top_percent: top N percent for when percent_or_count is "percent". Mutually exclusive with top_count.
• top_count: top N for when percent_or_count is "count". Mutually exclusive with top_percent.

Output

A signature attached via associated object security to the job.

Outlier Analysis (UpDownGenes)

Arguments

TODO

Output

TODO

NOTE: no gene set will be created when there are 0 outliers.

CBioUpdateData (UpdateCbioData)

Import data into local copy of cBioPortal.

This job exports data from mongo and runs Java code from cbioportal.org to load expression data clinical data from MedBook to cBioPortal.

Arguments

• sample_group_id: id of the sample group that defines which expression data will be loaded
• form_id: id of the clinical form for sample level data, sample_labels in this form must match sample labels from the sample_group defined above.
• patient_form_id: id of the clinical form with patient level attributes.

See here for Cbioportal importer tool documentation.

Output

Creates a log file of the resulting run and stores it in a blob in mongo.

Steps to adding a new tools

1. create a feature branch in git
2. look at an example job
3. create a new class
4. Add new class and its args to primary-collections repo collections/Jobs.js
5. add adapters (importers and exporters) to convert from MedBook objects to files that tools understand and store check them into external-tools
6. add external code to external-tools repo (or mechanism to install it)
7. add pointers to external code in your personal settings.json for use while testing

• also add these pointers in the MedBook main repo docker-compose.yml METEOR_SETTINGS environment variable

6. add gui to appropriate MedBook app, that initiates job by inserting into jobs collection for example:

Jobs.insert({
  name: "UpDownGenes",
  status: "waiting",
  user_id: user._id,
  collaborations: [ user.personalCollaboration() ],
  args: { ... }
});

7. read errors from jobs.error_description