A consideration of publication-derived immune-related associations in Coronavirus and related lung damaging diseases

Nophar Geifman; Anthony D Whetton

One rich, readily available source of immune-and disease-related knowledge is the corpus of published scientific research. Within research publications, a copious amount of diseaserelated trends are captured; these can be freely extracted from PubMed records [10] . We have previously developed a framework for extracting disease-immune relationships from PubMed abstracts by relying on the linking of such abstracts to concepts from Medical Subject Headings (MeSH) [11] . MeSH is the National Library of Medicine's controlled vocabulary thesaurus, consisting of terms (naming and descriptors) within a hierarchical structure; they are used for indexing MEDLINE PubMed publications. MeSH descriptors associated with each MEDLINE citation are manually assigned and provide a straightforward, and useful, knowledge resource. Numerous works using concept cooccurrences in biomedical texts or in associated MeSH terms have shown the utility of MeSH in capturing biomedical knowledge [12] [13] [14] [15] [16] [17] [18] . Our own assessment of the ability of MeSH descriptors associated within the same PubMed record to represent a true (meaningful and feasible) relationship between the terms has shown that co-occurrence of MeSH descriptors linked to any given PubMed record are a good source for mining dependencies between different types of biomedical entities [11] .