Issues with gene data

[ljpearlman]

I found several genes with links to both human and mouse species. The tables below have the species associated with those genes; the last column is the list of tables that gene/species was found in ("dataset_organism" is the join path gene -> dataset_gene -> dataset -> dataset_organism).

One MGI gene with both human and mouse data; something's probably wrong with the "human" data entries:

gene id	gene name	species	tables
MGI:99829	Runx2	Homo sapiens	dataset_organism
MGI:99829	Runx2	Mus musculus	gene_summary, dataset_organism, biosample

Three HGNC genes with both human and mouse data; something's probably wrong with the "mouse" data entries:

gene id	gene name	species	tables
HGNC:12660	VAX1	Homo sapiens	dataset_organism
HGNC:12660	VAX1	Mus musculus	dataset_organism
HGNC:6121	IRF6	Homo sapiens	dataset_organism
HGNC:6121	IRF6	Mus musculus	gene_summary, dataset_organism
HGNC:7391	MSX1	Homo sapiens	gene_summary, dataset_organism
HGNC:7391	MSX1	Mus musculus	gene_summary, dataset_organism

A bunch of Facebase-defined genes with both human and mouse data; I have no suggestions for these:

gene id	gene name	species	tables
FACEBASE:1-4SAG	Fgfr1	Homo sapiens	dataset_organism
FACEBASE:1-4SAG	Fgfr1	Mus musculus	gene_summary, dataset_organism
FACEBASE:1-4SAJ	Fgfr2	Homo sapiens	gene_summary, dataset_organism, clinical_assay
FACEBASE:1-4SAJ	Fgfr2	Mus musculus	gene_summary, dataset_organism, biosample
FACEBASE:1-4SBW	Tgfb3	Homo sapiens	gene_summary, dataset_organism
FACEBASE:1-4SBW	Tgfb3	Mus musculus	gene_summary, dataset_organism
FACEBASE:1-4SBY	Wnt5a	Homo sapiens	dataset_organism
FACEBASE:1-4SBY	Wnt5a	Mus musculus	dataset_organism, gene_summary
FACEBASE:1-4SC2	ABCA4	Homo sapiens	dataset_organism
FACEBASE:1-4SC2	ABCA4	Mus musculus	dataset_organism
FACEBASE:1-4SCE	NOG	Homo sapiens	dataset_organism
FACEBASE:1-4SCE	NOG	Mus musculus	dataset_organism
FACEBASE:1-4SD0	BMP4	Homo sapiens	gene_summary, dataset_organism
FACEBASE:1-4SD0	BMP4	Mus musculus	dataset_organism, gene_summary
FACEBASE:1-QK1M	FOXE1	Homo sapiens	dataset_organism
FACEBASE:1-QK1M	FOXE1	Mus musculus	dataset_organism
FACEBASE:1-QK1T	WNT9B	Homo sapiens	dataset_organism
FACEBASE:1-QK1T	WNT9B	Mus musculus	dataset_organism

[robes]

Seems like there are IDs for the BMP4 gene

• BMP4 (human) https://www.ncbi.nlm.nih.gov/gene?Db=gene&Cmd=DetailsSearch&Term=652
• Bmp4 (mouse) https://www.ncbi.nlm.nih.gov/gene/12159
• RUNX2 (human) https://www.ncbi.nlm.nih.gov/gene/860
• Runx2 (mouse) https://www.ncbi.nlm.nih.gov/gene/12393

If this is what we need to track down, I can pull the Gene ID out so that we can map from our existing terms.

[ljpearlman]

For that third group, it's possible that setting dataset_gene based on the species in dataset_organism will work; I think I just assumed there would be multiple organisms for each dataset because the dataset_organism table is many-to-many, but in practice that doesn't seem to be the case for most of them.

[ljpearlman]

This is done on fb-dev, except for the "has data" column (I have some questions about that). There's an ER diagram of the gene-related tables here.

Gene values in the data were translated like this:
For non-FACEBASE: (e.g., MGI or HGNC) gene ids, if there's a translation to an NCBI gene, then {source id, ncbi gene's species} is mapped to that gene.
For FACEBASE gene ids, if {source name or synonym, ncbi gene's species} has a unique case-insensitive match to an NCBI gene, then {source id, target species} is mapped to that gene.

All but two of the alternate_id values in the gene table were empty. I verified that the mappings came up with the same values (one gene had a uniprot id as its value, but that id mapped to the right NCBI gene).

There were 16 genes (affecting 14 datasets) that didn't map this way - two HGNC genes that appeared in mouse gene_summary records, one MGI gene that appeared in a human dataset (the only association of a species to that dataset appears to be a dataset_organism row), 8 FACEBASE genes that didn't match any names or ids, and 5 FACEBASE genes that mapped to multiple NCBI genes.

Gene ID	Name	Species in data	Problems	Matching NCBI IDs	Datasets (or gene summaries)
FACEBASE:1-4S9P	Col1	NCBITaxon:10090	no unique match	NCBI_Gene:12842, NCBI_Gene:12843	1-X5F8, 8C4
FACEBASE:1-4SAW	Hand2	NCBITaxon:10090	no unique match	NCBI_Gene:102636514, NCBI_Gene:15111	1-X5K0, 8DC
FACEBASE:1-4SB2	osterix/sp7	NCBITaxon:7955	no unique match	NULL	VHT
FACEBASE:1-4SCJ	panTro4	NCBITaxon:10090	no unique match	NULL	1-50KP
FACEBASE:1-QK2C	SMC2	NCBITaxon:9606	no unique match	NCBI_Gene:10592, NCBI_Gene:83452	1-JVTJ
FACEBASE:1-QK2R	ARNT	NCBITaxon:9606	no unique match	NCBI_Gene:375056, NCBI_Gene:405	1-JVTJ
FACEBASE:1-WKYJ	MIR124-3	NCBITaxon:10090	no unique match	NULL	1-HW9A
FACEBASE:1-WKYP	mirlet7a-5	NCBITaxon:10090	no unique match	NULL	1-HW9A
FACEBASE:1-WKYT	mirlet7b-5	NCBITaxon:10090	no unique match	NULL	1-HW9A
FACEBASE:1-WKZ0	mirlet7c-5	NCBITaxon:10090	no unique match	NULL	1-HW9A
FACEBASE:1-WKZ4	mirlet7d-5	NCBITaxon:10090	no unique match	NULL	1-HW9A
FACEBASE:1-Y7WW	Ror2	NCBITaxon:10090	no unique match	NCBI_Gene:19883, NCBI_Gene:26564	1-Y7W6, 1-YQK8
FACEBASE:1-YQW8	Clathrin	NCBITaxon:10090	no unique match	NULL	1-YQW0, 2-1QKP
HGNC:6121	IRF6	NCBITaxon:10090	no unique match, ontology/species mismatch	NULL	8DG
HGNC:7391	MSX1	NCBITaxon:10090	no unique match, ontology/species mismatch	NULL	8CC
MGI:99829	Runx2	NCBITaxon:9606	ontology/species mismatch	NULL	1-JVTJ

[robes]

I gather the issues with matching are mostly due to species mismatches. E.g., mirlet7a-N seem to be mostly issues of a zebrafish gene associated with a mouse dataset.

'panTro4' is just not a gene name. I can only guess the contributor mistook that field for gene assembly. I can clean that up in the production database. [DONE]

How is 'has_data' populated: A trigger, or some out of band process? Could we just filter on the Dataset and Biosample associations/references on 'has value' in the facet picker?