kgraph - Knowledge Graphs Constructions and Visualizations
Knowledge graphs enable to efficiently visualize and gain
insights into large-scale data analysis results, as p-values
from multiple studies or embedding data matrices. The usual
workflow is a user providing a data frame of association
studies results and specifying target nodes, e.g. phenotypes,
to visualize. The knowledge graph then shows all the features
which are significantly associated with the phenotype, with the
edges being proportional to the association scores. As the user
adds several target nodes and grouping information about the
nodes such as biological pathways, the construction of such
graphs soon becomes complex. The 'kgraph' package aims to
enable users to easily build such knowledge graphs, and
provides two main features: first, to enable building a
knowledge graph based on a data frame of concepts
relationships, be it p-values or cosine similarities; second,
to enable determining an appropriate cut-off on cosine
similarities from a complete embedding matrix, to enable the
building of a knowledge graph directly from an embedding
matrix. The 'kgraph' package provides several display, layout
and cut-off options, and has already proven useful to
researchers to enable them to visualize large sets of p-value
associations with various phenotypes, and to quickly be able to
visualize embedding results. Two example datasets are provided
to demonstrate these behaviors, and several live 'shiny'
applications are hosted by the CELEHS laboratory and Parse
Health, as the KESER Mental Health application
<https://keser-mental-health.parse-health.org/> based on Hong
C. (2021) <doi:10.1038/s41746-021-00519-z>. Additionally,
'kgraph' provides efficient methods to compute co-occurrence
matrices, pointwise mutual information (PMI) and singular value
decomposition (SVD) embeddings. In the biomedical and clinical
settings, one challenge is the huge size of databases, e.g.
when analyzing data of millions of patients over tens of years.
To address this, this package provides functions to efficiently
compute monthly co-occurrence matrices, which is the
computational bottleneck of the analysis, by using the
'RcppAlgos' package and sparse matrices. Furthermore, the
functions can be called on 'SQL' databases, enabling the
computation of co-occurrence matrices of tens of gigabytes of
data.
