Genetics Variant page review

prashantuniyal02 commented 3 weeks ago

General

Displaying numeric data:
- use numeric option in appropriate table columns (right-align, monospace font).
- format large integers with commas wherever appropriate
- round decimals wherever appropriate
Fix sorting where possibility of NA in column - comparator function should return 0 when both values nullish; nullish values should be larger/smaller than all other values dependent on intuitive sort direction for column

Variant page

Header:

Widget order?

[ ] In silico (V) > VEP (V2G) > ClinVar (V2D) > Uniprot (V2D) > GWAS (V2L2D) > molQTL (V2G) > Pharmacogenetics (V2D2Dr)

Metadata:

~~most severe consequence gene(s)?~~ [May be in 2.0]
~~Allelic frequency. Visualisation is odd. Numbers are relevant even on a log scale. Shall we be able to change from natural to log scale? Toggle?~~

Overall about widgets:

Save horizontal space [review data & number of rows]
Make sure top 25 rows is not an issue anywhere - pagination issues
~~Should we reuse the concept from Genetics of representing "V" "V-G" "V-D" in the top-right of the widgets? [not to be used in genetics widgets]~~
Categorical filter of columns would help in many places (Table component work). e.g. Affected tissue / cell etc.

In silico predictors:

Description: [to be discussed offline]
Tooltip for methods
Format methods name
Ideas?
color ranges would need to be based on some information - @DSuveges to talk to FE #3503

Variant effect predictor:

Description: [to be discussed offline]
[ ] Compress Aminoacid change + Coding Change + Uniprot accession columns. When aminoacid change is not NA show within the predicted consequence column: missense variant [R121Q]^? Tooltip: Uniprot accession: [P19438](Link to uniprot using identifiers.org)

ClinVar: good

Uniprot variants: good

GWAS Credible sets and molQTL credible sets (these changes also apply to the GWAS credible sets widget in the study page):

Description: [to be discussed offline]
[ ] Confidence (star system) #3605
[ ] Enhance [view] button UI. Title "View" and chavron?
[ ] LD column? Drop
[ ] Top L2G - (update to match the API )
[ ] L2G score (continuous filled bar)
[ ] Allow sorting by numeric columns (incl. posterior probability) (default p-value)
[ ] Affected tissue/cell: to show actual tissue label (not the ID) (e.g. blood plasma)
[ ] molQTL source : Source [Open Targets](link to Open targets homepage)
[ ] To add the condition to molQTL table

PheWAS plot:

to be scoped

buniello commented 3 weeks ago

Adding these points from #3597

VEP

Column headers need clearer definitions for better understanding

When downloading VEP data, users request: 1. Uniprot links 2. Chromosome and position for genes

In-silico predictors

Suggestion: adding hyperlinks or hover text to explain prediction methods. For example, SIFT=0 means "deleterious," which is opposite to PolyPhen where a score closer to 1 indicates deleterious

d0choa commented 2 weeks ago

@chinmehta one extra bit:

[ ] the default order of the VEP widget it would be by the column transcriptIndex (ascending). This column is not shown in the table, and we don't want it shown in the table (as it is derived from the combination of 2 existing columns). It means the user can not return to the original sorting if they start resorting the table, but we think we are OK with that.

d0choa commented 1 week ago

Another small request:

In the VEP widget, we would like to expand the query to include the biotype

query VariantEffectPredictorQuery($variantId: String!) {
  variant(variantId: $variantId){
    id
    transcriptConsequences {
      variantConsequences {
        id
        label
      }
      aminoAcidChange
      uniprotAccessions
      codons
      distanceFromFootprint
      distanceFromTss
      target {
        id
        approvedSymbol
        biotype #### New column
      }
      impact
      consequenceScore
      transcriptIndex
      transcriptId
      lofteePrediction
      siftPrediction
      polyphenPrediction
    }
    referenceAllele
    alternateAllele
  }
}

Using the new response, we would like to be able to tag in the Gene column all genes that have biotype == protein_coding. All the ones displaying different biotypes will not be tagged.

The most immediate UI I could think of is a chip displaying protein coding after the gene name and the tooltip. Probably an unfilled chip is more appropriate than the filled one, as we don't want to bring all the attention to it.

I'll gather more feedback from @buniello and others next week, but this should do the job.

d0choa commented 1 week ago

There is also a bug in the pharmacogenetics widget.

The next page shows no rows, but the API has some reasonable responses.

https://deploy-preview-524--ot-platform-partner.netlify.app/variant/7_117642566_G_A

There are some suspicious nulls in the phenotypeFromSourceId column that perhaps might cause this issue:

"pharmacogenomics": [
        {
          "genotypeId": "7_117642566_G_A,G",
          "isDirectTarget": false,
          "drugs": [
            {
              "drugFromSource": "ataluren",
              "drugId": "CHEMBL256997"
            }
          ],
          "phenotypeFromSourceId": null,
          "genotypeAnnotationText": "Patients with the rs77010898 AG genotype and cystic fibrosis may have improvement in chloride transport when treated with ataluren as compared to patients with the GG genotype. Randomized clinical trials found improvement in chloride transport, but did not find evidence for improved pulmonary function after 2 weeks of treatment. Other genetic and clinical factors may also influence changes in chloride transport and improvement of pulmonary symptoms in patients with cystic fibrosis.",
          "phenotypeText": "improvement in chloride transport",
          "pgxCategory": "other",
          "evidenceLevel": "3",
          "studyId": "1447954397",
          "literature": [
            "21233271",
            "18722008",
            "24836205",
            "20622033"
          ]
        },

Feel free to branch it out to a different ticket if significant

d0choa commented 1 week ago

One thing to try here (probably in separate PR) is to use a header of 2 rows in the credible set widgets.

For GWAS credible set:

2nd row header	1st row header
More details
Lead variant
P-value
Beta
Trait	Study
Diseases	Study
Identifier	Study
Posterior probability	Fine-mapping
Method	Fine-mapping
Confidence	Fine-mapping
Credible set size	Fine-mapping
Top gene	Locus2gene
Score	Locus2gene

For molQTL credible set:

2nd row header	1st row header
More details
Lead variant
P-value
Beta
Type	Study
Affected gene	Study
Affected tissue/cell	Study
Condition	Study
Posterior probability	Fine-mapping
Method	Fine-mapping
Confidence	Fine-mapping
Credible set size	Fine-mapping

opentargets / issues