Add ENFORMER annotations

[bschilder]

Would be great to access all genome-wide ENFORMER predictions via API. This should be possible since the predictions are shared as h5 files here. They're rather massive (14-42Gb each) but that should be mitigated by the h5 database format.

Alternatively, could extract the predictions on-the-fly from the pre-trained model. Usage examples here. But @AL-Murphy has mentioned that the pre-trained model they provide in the paper is not actually the one they describe in the paper.

[Al-Murphy]

Performance of available pre-trained model differs to what they show in the paper but the authors don't seem like they are going to do anything about this so I imagine the model available won't change.

What do you want these predictions for?

[bschilder]

Such a shame, but hopefully the annotations are at least generated by the model described in the paper?

The annotations could be used in at least two scenarios:

• Computing priors for functional fine-mapping.
• Conducting enrichment tests to validate fine-mapping SNPs (when not using the same annotations to perform functional fine-mapping).