Knowledge Bases¶

Comprior queries a knowledge base to retrieve relevant genes, gene association scores, or pathway information, which is then integrated into a prior knowledge feature selection approach
for the search, Comprior uses a) the class labels and b) alternative search terms that are provided in the config file (see alternative search terms parameter in configuration specification)

DisGeNET ¶

aggregates biological information from multiple sources (meta knowledge base, see sources description for original sources)
provides gene-disease, variant-disease, and variant-gene association scores for genes, disease, and variants (no pathway information) –> used for retrieving relevant genes and gene association scores
users can choose in the config whether to use DisGeNET’s gene-disease association (GDA) score, disease pleiotropy index, or disease specificity index (see their description for more information)
DisGeNET uses UMLS identifiers for the search, so Comprior internally first maps the search terms to their corresponding UMLS CUI (via the UMLS terminology web service) and forwards these to DisGeNET

provides biological information from multiple sources (meta knowledge base, see their documentation for original sources)
provides association gene-disease association scores (no pathway information) –> used for retrieving relevant genes and gene association scores

provides manually curated pathway information –> used for retrieving relevant genes, gene association scores, and pathway information
pathways are parsed into pypath.Network format

Comprior retrieves relevant genes from a set of pathways by selecting all their member genes and removing duplications

Comprior computes a gene association score \(s_i\) for a gene \(i\) from the sum of its degree percentile rank \(pr_{n,i}\) (= pathway genes are ranked by their number of in- and outgoing edges) for every pathway \(n\), normalized by \(P_i\) (=the overall number of pathways containing gene \(i\)): \(\frac{\sum_{n=1}^{|P_i|} pr_{n,i} if i \in p_n}{|P_i|}\)
this way, hub genes with a lot of interactions receive a higher score than genes at the outside rim of a pathway, becoming even more important if they have many interactions across multiple pathways

provides pathway information from multiple sources (meta knowledge base, see their sources for original sources)
relevant genes and gene association scores are currently retrieved the same way as KEGG
pathways are parsed into pypath.Network format