Motivation: Visualizing complex biological networks is a significant challenge, as traditional tools often struggle to represent data clearly and intuitively. Drawing inspiration from Piet Mondrian's abstract art style, the authors introduce Mondrian Map, a novel visualization tool that transforms the complexity of biological networks into a more organized, meaningful, and aesthetically appealing form. Results: In their case study of glioma progression, Mondrian Map reveals distinct pathway patterns across multiple patient profiles at different time points. Afterwards, the significance of these distinctive pathways in glioblastoma multiforme development is validated by recent literature. Mondrian Map's visually intuitive representation of complex biological networks enables researchers to easily identify crucial pathways and potential therapeutic targets. Moreover, The potential applications of Mondrian Map extend beyond biology, making it a versatile tool across domains. Availability and Implementation: The code was implemented in Python version 3.10 and is available through GitHub (github.com/fuad021/mondrian-map). The datasets used in this paper were retrieved from Glioma Longitudinal AnalySiS (GLASS) consortium.Abir, Fuad & Chen, Jake. (2024). Mondrian Map: Piet Mondrian's Abstraction for Differential Pathway Analysis. 10.1101/2024.04.11.589093.
In Fig. 3 B & E, the not differentially expressed PAGs (yellow) in aggressive profile exhibit a smaller area coverage due to their lower log2 of wFC compared to the up-regulated PAGs in the baseline. PAGs with IDs 2659, 2393, 2805, 3294, and 3196 show an increase in area in R2/TP compared to R1/TP. Notably, two down-regulated (blue) PAGs in R1/TP are absent in the R2/TP Mondrian Map, while two new up-regulated (red) PAGs emerge in R2/TP. Additionally, all the m-type relationships between the PAGs in R1/TP also persisted in R2/TP. In our retrieved data from the PAGER API, the PAG IDs have the common format of WAG00xxxx. Therefore, we have annotated the last 4 digits of PAG ID at the top-left corner of each rectangle for further analysis and validation. At this point, we have all the data retrieved and calculated for our visualization.
id: 3c53f19656cd24f365021c36e893a667 - page: 4
2.3. Qualitative Validation Notably, PAG with ID 3212 stands out in the top 10 PAGs for all six Mondrian Maps. In the aggressive R1/TP and R2/TP, it appears as a small, non-differentially expressed yellow rectangle whereas in the non-aggressive Mondrian Maps, the area increased, but it remains yellow. However, in the baseline Mondrian Maps, PAG 3212 is distinctly up-regulated, marked by the large red rectangles. Additionally, this PAG forms two m-type relationships with other PAGs in baseline and aggressive profiles, underscoring its significance for further analysis.
id: 8cadfed130a61c1f8ae2d518add88a7a - page: 4
To validate the patterns we identified by analyzing Mondrian Maps, we used a language model-based search engine called Perplexity AI (Perplexity AI, 2022). By querying the PAGER Web App (Yue et al., 2022) with specific PAG IDs, we obtained details about the pathways we were interested in. With this pathway information in hand, we then queried the search engine to explore any connections between the disease and these pathways. The Furthermore, all the PAGs in non-aggressive R1/TP Mondrian in Map are yellow, R2/TP Mondrian Map, PAG ID 2133, 2628, 2876, 2483 and 2783 are up regulated and the increment in area is in many folds. PAGs in lower-right portion of R1/TP and R2/TP for non-aggressive Mondrian Maps (Fig. 3 C & F) are labeled with their IDs as they i.e., not differentially expressed. However,
id: 7964ca7a4fb28c75dfb8d840210c2581 - page: 4
It is doi: ; this version posted April 15, 2024. made available under a CC-BY-NC 4.0 International license .
id: 5e1ba4e09ebc35e36b1e32235532885a - page: 5