Building a knowledge graph to enable precision medicine

Abstract

Developing personalized diagnostic strategies and targeted treatments requires a deep understanding of disease biology and the ability to dissect the relationship between molecular and genetic factors and their phenotypic consequences. However, such knowledge is fragmented across publications, non-standardized research repositories, and evolving ontologies describing various scales of biological organization between genotypes and clinical phenotypes. Here, we present PrimeKG, a precision medicine-oriented knowledge graph that provides a holistic view of diseases. PrimeKG integrates 20 high-quality resources to describe 17,080 diseases with 4,050,249 relationships representing ten major biological scales, including disease-associated protein perturbations, biological processes and pathways, anatomical and phenotypic scale, and the entire range of approved and experimental drugs with their therapeutic action, considerably expanding previous efforts in disease-rooted knowledge graphs. PrimeKG supports drug-disease prediction by including an abundance of ‘indications’, ‘contradictions’ and ‘off-label use’ edges, which are usually missing in other knowledge graphs. We accompany PrimeKG’s graph structure with text descriptions of clinical guidelines for drugs and diseases to enable multimodal analyses.

Few shot learning for phenotype-driven diagnosis of patients with rare genetic diseases

“…Additional associations of diseases, genes, and phenotypes since then may further improve SHEP-HERD's performance. To this end, the knowledge graph curation and processing approaches are fully reproducible, and the graph can be automatically updated as data resources evolve and new data become available [59]. Second, the still-undiagnosed UDN patients may be more challenging than the already-diagnosed ones SHEPHERD was tested on.…”

“…We create a comprehensive knowledge graph (KG) for rare disease diagnosis. We start with PrimeKG [59] and adapt it to the rare disease setting by removing drug-specific entities and relations and adding additional sources of the gene, phenotype, and disease relationships. The resulting rare disease KG contains seven node types (i.e., phenotype, protein, disease, pathway, molecular function (MF), cellular component (CC), and biological process (BP)) and 15 unique relation types (i.e., phenotype-protein, disease-phenotype (-) (indicating that disease does not have phenotype), disease-phenotype (+) (indicating that disease has phenotype), protein-pathway, disease-protein, protein-MF, protein-CC, protein-BP, BP-BP, MF-MF, CC-CC, phenotype-phenotype, protein-protein, disease-disease, pathway-pathway).…”

“…We perform all other preprocessing as in the original PrimeKG knowledge graph. For additional information about each data source and the harmonization process, refer to PrimeKG [59].…”

CascadER: Cross-Modal Cascading for Knowledge Graph Link Prediction

“…Currently, the prevailing approach is to learn vector representations of entities and relations, or knowledge graph embeddings (KGEs), and apply various vector composition functions to the embeddings to score candidate links [Ruffinelli et al, 2020]. While often effective at modeling structural and logical KG patterns [Sun et al, 2019], KGEs typically do not utilize the abundant textual information in KGs, even though auxiliary texts like entity descriptions can ameliorate KG sparsity [Chandak et al, 2022] and improve ranking accuracy [Xie et al, 2016].…”