How is the V2G score calculated?

Using rs4975758 as an example, the web interface gives the V2G score for the gene IRX4 as 0.199496222.

When I get the same data from variant_gene table, I do not see this value.

Documentation provides the code, so it looks like it is calculated on-the-fly.

Question, is it possible to recalculate that Overall V2G score using data available on GBQ?

1 Like

Using the weights from docs:

source_weights = {
    'vep': 1,
    'javierre2016': 0.33,
    'andersson2014': 0.33,
    'thurman2012': 0.33,
    'canonical_tss': 0.33,
    'gtex_v7': 0.66,
    'sun2018': 0.66

sum qtl_score_q and multiply with weight 0.66:
(0.7 + 0.4 + 0.4 + 0.1) * 0.66 = 1.056

distance_score multiply with weight 0.33:
0.000261506276150627 * 0.33 = 0.00008629707

weighted mean:
(1.056 + 0.00008629707) / (0.66 + 0.66 + 0.66 + 0.66 + 0.33) = 0.3555846118

But the number on the website is 0.199496221662468, what am I missing?

Can anyone please clarify?


Sorry for the slow reply, your question is on the radar of our backend team. We’ll let get back to you as soon as our capacity allows to address your question.


1 Like

@dsuveges any updates?

Hi @zx8754!

The V2G overall score is an aggregation of the different scores across sources and as you point out, it is not part of the variant_gene table. We pre-compute these values in order to display them in the FE. If you are curious in the score for a particular variant/gene, you can use the genesForVariant endpoint of the API to fetch the overallScore .

If you are interested in a larger set of variants, the scored V2G dataset can be derived from the V2G one we make available on BigQuery. To make so, the logic is available in our Genetics ETL here.

I hope that it is helpful! We are currently working towards harmonising the Genetics infrastructure with the one we have for the Platform, so we very much appreciate any comments that help us to make the data more useful.


1 Like