OncoKB Analysis in Study View

[zhx828]

Background: The Study View (e.g. MSK-IMPACT 2017 study) of cBioPortal provides a powerful interactive way for exploring clinical and genomic data in a study (cohort of samples).

OncoKB is a precision oncology knowledge base and contains information about the effects and treatment implications of specific cancer gene alterations.

We have been integrating OncoKB knowledge into the portal. You can see the integration in the Annotation column in both result's page and patient page.

All the annotations for pages mentioned above are in real-time. But the analysis in the study view is really challenging due to the massive genomic data of each study, especially in PanCancer Studies.

Goal: The overall goal of this project is to integrate OncoKB biological effect and clinical implications into cBioPortal Study View by storing pre-annotated results in the database. This includes both backend(JAVA) and frontend(JS/React) implementations.

• Backend goal
  • Create an additional column in mutation table annotation
  • Pre-annotated each mutation and store the annotation in the MySQL database (mutation table)
  • Implement database queries and web API endpoints to serve the frontend's needs
• Frontend goal
  • Create two pie charts: "Has drivers" and "highest therapeutic level" per sample.
    • Has known driver alterations
      • Yes: a sample with at least one Oncogenic, Likely Oncogenic, or Predicted Oncogenic mutations per oncokb
      • No: Likely Neutral, Inconclusive, Unknown
    • Highest Level of Sensitive Therapeutic Implications per sample
      • Level 1
      • Level 2A
      • Level 2B
      • Level 3A
      • Level 3B
      • Level 4
      • Oncogenic, no level -VUS
  • Create a stacked bar chart of levels in different cancer types.

Approach: Ideally, the implementation should be generic, ie. not OncoKB-specific, so that any driver and their annotation and tiers can be used.

Backend:

• Create an additional column and store the OncoKB annotation in the column store in JSON format. You can also count and groupby within json: https://stackoverflow.com/questions/37347443/mysql-json-datatype-count-and-groupby
• We need a script to annotate and reannotate all mutations in the database (currently OncoKB annotation is currently pulled from OncoKB API directly)
• Insert ONCOKB_HAS_DRIVER attribute into clinical_sample table based on the annotation for each mutation
• Develop web API to serve the frontend needs

Frontend:

• Implement the new two charts described above and in https://github.com/cBioPortal/cbioportal/issues/5212
• Implement an option to only use driver events for Mutated Genes and CNA Genes table.

Need skills: JAVA, MySQL, ReactJS, Javascript, HTML, CSS

Possible mentors: @zhx828

[AndrewCveklNYUClass]

Hi, very interested in working on this.

[kloun]

my name is Victor Kukshiev. I very interested in working on this project. I familiar with SQL, web and java...

[jjgao]

@nabeelmerchant1 GSoC is only open to undergraduate and graduate students.

[jjgao]

@nabeelmerchant1 Thanks for submitting your proposal. Please rest assured that we will review your proposal carefully together with proposals from others. But please understand that none of the submitted proposals are guaranteed to be accepted and therefore none is taken. (Hongxin and I don't really have the power to approve that.)

Also, thanks for sharing your draft here, but to be clear to other students, we don't recommend sharing drafts on GitHub.

[zhx828]

Add /oncokb/fetch POST for study view which is similar to /clinical-data-counts/fetch

Accepts two parameters

• attributes
• studyViewFilter

Possible attributes

• HAS_DRIVER
• HIGHEST_SENSITIVE_LEVEL

And in StudyViewFilter, we need to add

List<OncoKBDataFilter> oncokbFilters;
class OncoKBDataFilter {
    private String attributeId;
    private List<String> values;
}

The response will be similar to /clinical-data-counts/fetch as well

[
  {
    "attributeId": "string",
    "counts": [
      {
        "count": 0,
        "value": "string"
      }
    ]
  }
]

[kalletlak]

@zhx828 what would be the response?

[zhx828]

@kalletlak it will be similar to /clinical-data-counts/fetch. I updated the comment above.

[inodb]

@zhx828 @kalletlak I guess we can rewrite this a bit for GSoC2020?