The Harmonizome file is tab separated and has two header lines: the column
names (source, source_desc, source_id, target,
target_desc, target_id, weight) followed by a
descriptor row (GeneSym, NA, GeneID, Phenotype,
HPOID, NA, weight). The parser drops the descriptor row
and renames the columns to domain-specific names. Columns that carry no
information in this dataset (source_desc and the -666
placeholder in target_id) are not kept.
Gzip-compressed files (.gz) are read directly.
The resulting object can be linked to DrugBank target genes with
merge_drugbank_hpo.