Label Size | |
Label Size |
Nodes | |
Edges |
Nodes | |
Edges |
Several hundred clinical trials on COVID-19 were launched within a few weeks worldwide. Planned/ongoing/terminated trials are being tracked by WHO and physicians.
The provision of a synthetic view of these clinical trials is essential for the coordination of the research community: to explore new therapeutic targets while avoiding similar studies.
This is the objective of this mapping which, based on textual data mining of the metadata of these trials, allows to see at a glance all the avenues explored by current and past clinical trials, the combinations of treatments, their main categories and the expected or observed results.
What you see are the main categories of treatments and outcomes and for each of them, the most related treatments and outcomes are highlighted. Two terms are linked on the map if there is a statistical relation in the set of clinical trials between them.
These maps are based on database of randomized COVID studies provided by the WHO and pre-processed by an international collaboration of laboratories that filter, clean, enrich the dataset. We got it from Pr Isabelle Boutron (CRESS-UMR1153) on April 8 2020. The aim of this database is to inform about current studies and explore new therapeutic targets, avoiding similar studies. It synthetises more than 600 clinical trials arms on coronavirus.
These studies are all pre-recorded and therefore normally ethically validated; however, some have not yet begun.
As of April 8 2020, a total of 274 studies have been recorded. This database is a manual work so there may be some inconsistencies, it has been enriched with the treatment families.
These maps has been realized using Gargantext on April 11 2020. They aim at giving a bird eye view of all the randomized COVID studies mentionned in the WHO database. We did text-mining on the treatment description and outcome descriptions to infer the main categories of clinical trials and their possible complementarities.
The methodology is described hereWhen you click on a term, the most related terms are displayed in a tag cloud on the right along with the details of the clinical trials that mention the most of the selected terms.
In the Distributional maps (the first map to be loaded) links between terms capture a proximity measure that takes into account the profiles of interaction of each term with the others. It can infer relevant relations between terms even if their haven't co-occur in a clinical trial (second order proximity measure).
In the Conditional map, links between terms represent the conditionnal probability of having one terms knowing the other in a paper (its the confidence measure).
In each graph, the node size is a fonction of the pagerank of the term in the graph.
The mathematical formula of these proximity measure can be found in the Gargantext documentation.
This map uses a CNRS open source tool that has been diverted from its original use for visualization examples (ProjectExplorer). The display of documents associated with a selection on the map has a few bugs that we will try to correct if this way of presenting clinical trials proves useful for researchers and practitioners working on coronavirus.
here video