Semantic Web and linked data technologies garner significant interest and investment from the biomedical research community toward tackling the diverse challenges of heterogeneous biomedical data and knowledge integration. We have extracted schemas and vocabularies from more than 90 publicly available biomedical linked data graphs into an LSLOD schema graph and conduct an empirical meta-analysis to evaluate the extent of semantic heterogeneity across the LSLOD cloud. The total number of classes, object properties, data properties, and datatypes extracted and linked in the LSLOD schema graph were 57,748, 4,397, 8,447 and 24 respectively. Through an empirical analysis, we have found some overlap between the different schema elements, based on the way linked data publishers model the underlying data (i.e., some object properties in a given source are used as data properties in another linked data source).
The entire LSLOD schema graph can be downloaded at Figshare and explored in this website below. The findings of our research are currently under review in Nature Scientific Data and a preprint will be uploaded shortly on Arxiv.