There exists a large amount of annotation data that overlays the human genome. Many users want to ingest this data in order to join it against their database of genome variants. We want to generate a general schema that will cover the vast majority of these annotations. Example sources include ENCODE, UCSC genome browser, dbSNP/COSMIC/etc.
Much of the data come in several common text formats (e.g., BED, GFF), reviewed here: https://genome.ucsc.edu/FAQ/FAQformat.html
Below is the proposed schema for genome annotations. In addition to the schema, there will be ADAM tools to ingest the most common formats.
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25
Some annotations may be hierarchical. For example, ChIP-seq|sample1|TF2. How could you aggregate across the different “fields” in the annotation. In principle, these should all be in the
BED format allows for blocked annotations (e.g. exons). If we ingest this kind of data, should the solution be that each exon is its own annotation, with an additional “grouping” field that represents the gene name? Alternatively, you could instead have a nested annotation object. (Are self-referential types allowed? The ADAMAnnotation could have an array of ADAMAnnotation) in it.
Or should we just scrap this general annotation type and have a specific type for each of the popular formats (e.g., ADAMBED, ADAMGFF, etc.)?
Or do a Google-style monster annotation type that simply contains pretty much any field anyone would ever want. As ADAM evolves, we would simple continue to add more and more fields.