A Framework To Build A Causal Knowledge Graph for Chronic Diseases and Cancers By Discovering Semantic Associations from Biomedical Literature

Abstract

Extracting knowledge from disparate biomedical literature can play an important role in the discovery of disease mechanisms and remedial therapies. This paper explores a hybrid semantics-based knowledge synthesis and discovery methodology that integrates approaches from Literature Based Discovery (LBD), Systems Medicine, and Knowledge Graphs to analyze published biomedical literature and discover potential causal associations between risk factors and Non-Communicable Diseases (NCDs). This paper presents a knowledge synthesis and discovery framework to (a) mine biomedical literature to identify semantic associations between risk factors and NCDs, and (b) represent them as a knowledge graph that outlines the multi-causal associations between underlying risk factors and NCDs. We employ a novel ranking algorithm that considers direct and indirect relation-based methods, augmented by semantic relatedness, to discover causal associations between risk factors and a targeted condition—in this case breast cancer. The novelty of our work is the use of breast cancer-specific embeddings in combination with graph-based metrics to quantitatively evaluate semantic association based on causality. We evaluate the performance of our breast cancer-specific word embedding model by utilizing information retrieval methods and manually curated breast cancer relations. Results confirm that (a) our cancer-specific word embedding model out-performs non-disease-specific models with respect to retrieval of breast cancer relations, and (b) our method generates valid causal knowledge about causal risk and protective factors related to breast cancer. Our present study focuses on breast cancer, however, our method is adaptable to other NCDs.