Load and expose through API L2G predictions

[d0choa]

Within the data-freezes we are producing a dataset named: locus_to_gene_predictions

❯ gsutil du -sh gs://ot_orchestration/releases/24.10_freeze3/locus_to_gene_predictions/
726.59 MiB   gs://ot_orchestration/releases/24.10_freeze3/locus_to_gene_predictions

This Dataset contains all L2G predictions in 3 columns:

+----------------------------------+-----------------+-------------+
| studyLocusId                     | geneId          |       score |
|----------------------------------+-----------------+-------------|
| bd62bdaacefab55429a796fee1f1625f | ENSG00000235016 | 0.00608669  |
| -3667170787811173525             | ENSG00000251840 | 0.101783    |
| -2434565126434848277             | ENSG00000149294 | 0.0207067   |
| 5659704022772720604              | ENSG00000216990 | 0.0261894   |
| -2208135635528252608             | ENSG00000288879 | 0.0281841   |
| 6906907568713546539              | ENSG00000213304 | 0.00156307  |
| 1db879a31caa452788438f8a45fb7111 | ENSG00000132631 | 0.0017865   |

We would like to add this dataset to OpenSearch. The schema can be found in gentropy. There should be no nulls, and studyLocusId + geneId should define unique rows.

Within the API, we would like to add another object to the credible set object named L2Gpredictions (or whatever looks consistent), which would return a list of the resolved entities geneId -> target and the score for each prediction.

If it doesn't add much complication, we would like the API to sort the score by descending order, so the FE can easily take the top L2G from the list.

For any issues with the data, contact @ireneisdoomed.

@buniello / @gjmcn this will rescue the top L2G functionality but it would require minor adjustments

@buniello / @addramir / @DSuveges This will potentially unblock an L2G widget in credible set page, but at the moment, it will be a little bit vague until we work out how to present the L2G features.