Immune-centric network of cytokines and cells in disease context identified by computational mining of PubMed

Abstract

Cytokines are signaling molecules secreted and sensed by immune and other cell types, enabling dynamic intercellular communication. Although a vast amount of data on these interactions exists, this information is not compiled, integrated or easily searchable. Here we report immuneXpresso, a text-mining engine that structures and standardizes knowledge of immune intercellular communication. We applied immuneXpresso to PubMed to identify relationships between 340 cell types and 140 cytokines across thousands of diseases. The method is able to distinguish between incoming and outgoing interactions, and it includes the effect of the interaction and the cellular function involved. These factors are assigned a confidence score and linked to the disease. By leveraging the breadth of this network, we predicted and experimentally verified previously unappreciated cell-cytokine interactions. We also built a global immune-centric view of diseases and used it to predict cytokine-disease associations. This standardized knowledgebase (http://www.immunexpresso.org) opens up new directions for interpretation of immune data and model-driven systems immunology.

Extracting T Cell Function and Differentiation Characteristics from the Biomedical Literature

“…A single link in this case is only included when found in at least 10 distinct documents to ensure that the link is more likely to be valid. This is comparable to Figure 2c of [1] where the Pearson correlation between the number of links was found to be .86, very similar to the value of .93 found in this study.…”

“…Supplementary Figure 10 shows precision scores for various thresholds used to make this determination and Supplementary Figure 11 shows high probability predictions for each relation, to better clarify how ideal relation candidates appear. For further validation of T cell subtype and cytokine relations extracted, we also compared our results to those of [1], which are summarized in Figure 3 as well as Figure 4 . Figure 3 shows that the number of documents associated with each relation varies based on the source (ours vs theirs) but that there is a significant level of correlation.…”

“…Two separate corpora were collected for this study. The first (aka "Dev Corpus") was based on a keyword query passed to the Entrez API in an effort to collect a smaller, high relevance corpus 1…”

“…Entity tagging for cytokines, transcription factors, and cell types was applied using known symbol and alias lookup tables . Cytokine names and aliases were sourced primarily from [1] 2 (via Cytokine Registry ) and [8]. Transcription Factor names as well as associated aliases were integrated primarily from [8] with a large number of transcription factor aliases also provided by [9].…”

“…All of the links extracted relate exclusively to T cell subtypes and include cytokines that either induce differentiation (hereinafter referred to as "inducing cytokines") of these subtypes or are secreted by them ("secreted cytokines"), as well as the transcription factors involved in the path along any developmental trajectory ("inducing transcription factors"). This former type of link, between T cell subtypes and secreted or inducing cytokines, was quantified over a large corpus and shared first in the immuneXpresso [1] (iX) database, which also includes a much larger breadth of cell types and covered the entirety of PubMed abstracts available at the time. A primary goal in this study is to be able to recapitulate much of this data while utilizing more expressive language models trained with less manual annotation effort, a capability conferred by recent advancements in weak supervision.…”