Visual output of the CNG

The CNG user interface shows the results of an ongoing or finished neural network training process. These are documented in HTML files which include textual information and plots. The subsequently generated index file is the starting point for the visualization. It shows all settings of the training process as well as an overview of all generated networks. Each generated network is also described and visualized in its own HTML file. The binary files of the training process and the single networks are automatically created with the HTML report. The index HTML file also shows the differences of the categorized sequences of the networks. This is done by generating a distance matrix of all generated networks. A value of 1 means that no sequences can be found in both categories, whereas a value of 0 means that all sequences of the smaller category are included in the larger category. This distance matrix is visualized as a 2D plot using the “symmetric SMACOF” multidimensional scaling method [95]. Additionally, the index file also shows whether a correlation between each of the input data of the categorized sequences of each particular network was detected using Spearman‘s correlation coefficient. The HTML files describing the single networks show a scatterplot of the input data, boxplots of the single data, the categorized sequences as well as a table containing the all included CREs and the genes in which they occur. The gene identifiers are provided in separate text files. These gene identifier lists allow subsequent analyses, such as GO analyses.