Fig 1.
(A) Overview of the proposed approach. (B) The Log likelihood score across different number of topics. (C) The Log likelihood score across different iterations.
Fig 2.
(A) Distribution of diseases and genes across 160 optimal disease topics. (B) The heatmap of cosine similarity for top 10 topics presented at disease level. (C) The heatmap of cosine similarity for top 10 topics presented at gene level. (D) Overall Distribution of 146 LDA Topics on 19 Human Disease Network Categories in Goh et al.
Table 1.
Topics with the most diseases mapped on Human Disease Network Categories.
Table 2.
Top 10 LDA topics containing most OMIM disease-gene associations.
Fig 3.
(A) Top 10 topics and their corresponding top 5 diseases based on probabilities. (B) Top 10 topics and their corresponding top 5 genes based on probabilities. For both figures, color blue, red, green, purple, and cyan represent top 1 to 5 diseases/genes respectively.
Fig 4.
(A) The precision recall curve for top ten topics annotated by three independent disease ontologies. (B) Area under curve (AUC) score for top 10 topics using three independent disease ontologies.
Table 3.
Statistics of top ten disease topics.
Table 4.
A list of enriched diseases and disorders associated with genes in the AD association network.