Designing a public CRISPR widget

[buniello]

As we make progress with processing and interpreting CRISPR results from

Brain CRISPR #2821 BIOGRID #2920

we need to design a public CRISPR widget for the 23.06 release - keeping in mind that the PPP will host both a public and a pre-publication widget.

This issue will be completed with widget specifications once the schema for these datasets is finalised, as discussed with @DSuveges.

Evidence schema (example from Brain CRISPR):

{
  "studyId": "Glutamatergic Neuron-CellRox-CRISPRi",
  "datatypeId": "sysbio",
  "datasourceId": "crispr_screen",
  "projectId": "crispr_brain",
  "studyOverview": "CRISPRi FACS screen for reactive oxygen species in human iPSC-derived glutamatergic neurons.",
  "experiment": "WT CRISPRi-iPSCs were transduced with genome-wide CRISPRi-v2 sgRNA libraries. Two days after infection, the cells were selected for lentiviral integration using puromycin (1 g/mL) for 3 days as the cultures were expanded for the screens.",
  "cellType": "Glutamatergic Neuron",
  "contrast": null,
  "diseaseFromSource": "neurodegenerative disease, ageing",
  "diseaseFromSourceMappedId": "EFO_0005772",
  "targetFromSource": "MIS12",
  "pValue": 0.1350578615532632,
  "statisticalTestTail": "lower tail",
  "resourceScore": -3.5780208400785147
}

[DSuveges]

@buniello for your interest: I have reviewed the ot_crispr dataset schema and made a few changes to make sure the two are more similar. Now it looks like this:

{
  "studyId": "Glutamatergic Neuron-Survival-CRISPRi",
  "datatypeId": "sysbio",
  "datasourceId": "crispr_screen",
  "projectId": "crispr_brain",
  "crisprScreenLibrary": "Genome-wide CRISPRi-v2",
  "studyOverview": "CRISPRi survival screen in human iPSC-derived glutamatergic neurons (standard medium).",
  "cellType": "Glutamatergic Neuron",
  "literature": [
    "34031600"
  ],
  "diseaseFromSource": "essential genes / neurodegenerative disease",
  "diseaseFromSourceMappedId": "EFO_0005772",
  "contrast": "Survival",
  "targetFromSourceId": "SOD2",
  "resourceScore": 0.0029283011705045,
  "statisticalTestTail": "lower tail",
  "log2FoldChangeValue": -24.05642773885583
}

• The type of the datasource, consistent with projectscore, is sysbio.
• The identifier of the datasource is crispr_screen that would pool different sources eg. Brain CRISPR, BioGRID ORCS etc. (similar to gene burden)
• The identifier of the project tells which source the screen was integrated from. In this case crispr_brain. This would allow both the user and the frontend to connect the piece of evidence.
• The studyId identifies the screen in the project. In this case it allows the frontend to generate the link to crisprBrain page: Glutamatergic Neuron-Survival-CRISPRi
• There are other screen specific fields to contextualise the experiment: crisprScreenLibrary, cellType
• Contrast is either curated and pulled from the disease mapping file or pulled from the Phenotype column in the study table. (which is always the case at the moment, as we have no curation for any of the studies).
• diseaseFromSource and diseaseFromSourceMappedId are all curated by our experts. The mapped identifier is the EFO id. (if the diseaseFromSource could be mapped to multiple EFO terms, the evidence is exploded).
• targetFromSrouce - Hits are coming from the already annotated gene lists from CRISPR Brain. Where the annotation for the genes is either positive or negative hit.
• statisticalTestTail is inferred by the annotation positive or negative hit
• resourceScore is the p-value of the gene
• log2FoldChangeValue this is the phenotype score. I'm not 100% sure if this is indeed log2 fold change.

[buniello]

First draft of CRISPR Screens(public) widget - please see data schema above for reference

CS CRISPR Screens CRISPR knockout screens from public CRISPR datasources, associating "targetFromSourceId" and CRISPR results. Sources: "projectId" (link) e.g. CRISPRBrain (will be a comma separated list when data from more sources will be integrated)

Table Reported disease: "diseaseFromSource" (Link to “diseaseFromSourceMappedId” disease page - can be multiple links for multiple mappings) Study Identifier: “studyId" (Link link to CRISPRBrain study page: e.g. Glutamatergic Neuron-Survival-CRISPRi) - will have different identifiers from different sources when integrating more sources Contrast/Study overview: "contrast”/ "studyOverview" (tooltip: SCREEN LIBRARY: "crisprScreenLibrary") Cell type: "cellType" Log2 fold change: "log2FoldChangeValue" Significance: "resourceScore" (tooltip STATISTICAL TEST TAIL: "statisticalTestTail”) Source: "projectId" (link to source e.g. CRISPRBrain) - we plan to integrate data from multiple sources with time, so this link can vary depending from the source Publication: "literature" (link to EPMC similar to OT Genetics widget)

[buniello]

@LucaFumis: "datatypeId": "sysbio" Position of new data source on the "Association on the Fly" page: before Project Score (first data type for the sysbio aggregation)

[buniello]

Thanks @LucaFumis for the first version of the CRISPR Screens widget - it looks good! Just a few minor comments, as already discussed on Slack - Contrast / Study overview : delete full stop at the end log2 fold change: format (i.e. trim) source: format it to CRISPBrain (same as link in the header) diseaseFromSource: when the record returns "essential genes / neurodegenerative disease" - visualise as neurodegenerative disease (essential genes) or neurodegenerative disease (tooltip: essential genes)

[LucaFumis]

Thank you @buniello - I've updated the preview with the changes we've discussed today and the new datasource: looks to be working correctly in both AOTF and classic associations. I have to fix the table download.

[buniello]

In a second iteration (following releases) the CRISPRBrain widget/data need some cleaning-up within the reported disease field. Some examples of datasets needing cleaning up within the reported disease field - by expanding the reported diseases one by one - are:

https://platform.dev.opentargets.xyz/evidence/ENSG00000132639/GO_0007568 https://platform.dev.opentargets.xyz/evidence/ENSG00000176884/MONDO_0005180

This is for myself and Daniel to remember.