README.md

In xinxiong0238/CUInetwork: Network visualization for NLP and codified concepts

CUInetwork

The goal of CUInetwork is to generate an interactive shiny app that help visualize the network of codified &/ NLP concepts.

Installation

To install the latest development version from GitHub:

install.packages("remotes")
remotes::install_github("xinxiong0238/CUInetwork")

If errors like “Error: (converted from warning) package ‘shiny’ was built under R version 4.0.5” occur, please set:

Sys.setenv("R_REMOTES_NO_ERRORS_FROM_WARNINGS" = "true")

You can refer to this website to see more discussion on that problem.

Example

This is a basic example which shows you how to run the CUInetwork app. Remember you need to get access to the data and save it to your local computer. Remember in order to guarantee some dependencies are loaded, you must use library(CUInetwork) beforehand, instead of directly running CUInetwork::run_app(data_path)!

library(CUInetwork)
data_path = "the path to the app data (RDS format)"
run_app(data_path)

App instructions

To have a general understanding on how to get the CUInetwork app work, we show some main steps here. If you want to know some detailed app functionality, please refer to the ‘Get started’ tab.

Step 1: choose your center node(s)

First of all, you need to specify your center node by clicking a row in the Possible inputs table containing in the dropdown button. Multiple inputs are acceptable. You can also use the search bar on the top or the filtering bar for each column (id, terms, group) to spot on your targets. A second round of searching will not mask the previous selection, which will be kept unless you de-select the row or use the “Unselect” button on the navigation bar. Once some rows are selected, the app will automatically display them as selected CUIs or nonCUI codes on the left.

Step 2: show the connected network

When you select your interested center node(s), simply click the green “Show” button to take a look at the connected network. Remember only nodes directly linked to your center node(s) will be shown. For example the following network only draws CUIs or codified concepts directly connected to phecode 174.1 or phecode 174.2. In terms of nodes that have indirect connection like A \<-> intermediate node \<-> center node, they are not in the interests of this app.

Step 3: explore network information

Main panel

In general, there are CUIs (or NLP nodes) and nonCUIs (codified concepts) as two major classes of nodes, which we use circles and triangles to represent. For CUIs, By hovering your mouse onto any node, you will see the basic information including node ID and group name.

If hovering on an edge, a tooltip will show up with the connected node information and the cosine similarity. However, this is not recommended when the center node is surrounded by too many nodes, which easily leads to mistouch.

You can use the nagivation buttons on the bottom to control the size of the network and move the center of the graph. The same funcionality can be achieved by scrolling up or down the mouse wheel or dragging any blank area in the main panel. Clicking the download button on the top right corner will start downloading a csv file including all edge information on the graph (five columns: center node id/label, connected node id/label and cosine similarity). The questionmark button opens an instruction window.

Popup window

When the app generates the connected network for your center node(s), you may want to dive into one specific node and take a further look. This node can be one of the previous center nodes, or it can be a non-center node but connected to a center. To achieve that, you just need to click on any node on the canvas and a popup window will then bring you more details (sunburst plot, static/interactive circular plot and data table).

Remember, if you click on a non-center node, there is a chance that the node is not in the interested list of the current app. If so, the popup window is not able to provide much more information. The “Possible inputs” table shows all nodes in the interested list.

xinxiong0238/CUInetwork documentation built on Dec. 23, 2021, 6:21 p.m.