Fig 1.
Visualization of the analysis workflow.
Fig 2.
Schematic visualization of a possible sequence of steps performed during the analysis.
Different combinations of the steps are possible depending on the data set and research question. The * indicates that stage is optionally and, depending on the data set and research question, might not be required to obtain a small and meaningful data set.
Fig 3.
Visualization of the workflow extracting characteristic genes.
Created with BioRender.com.
Fig 4.
Clustering of the Seurat object.
(A) Clustering by cell types and ages. (B) Clustering by age.
Fig 5.
Visualization of the selected PFA genes.
(A) UMAP plot of the relevant PFA genes. (B) Clustering according to DBSCAN. (C) Composition of the different clusters according to condition (OLD or YOUNG).
Fig 6.
HDBSCAN can be used as an alternative to DBSCAN.
(A) Clustering according to HDBSCAN. (B) Composition of the different clusters according to condition (OLD or YOUNG).
Fig 7.
Gene expression and condition of the five top-ranked results.
(A) SHAP values of cluster 5 (mostly old cells; compared to the young cells, the gene expression is upregulated in the old cells). (B) SHAP values of cluster 8 (mostly young cells; compared to the old cells, the gene expression is downregulated in the young cells). (C) The tree explanation visualizes the top-ranked result (in this showcase PLK3) and the clustering for PLK3 expressing samples. 17.0% of the PLK3 expressing cells are among the cells of cluster 8 (mostly young cells), and 83.0% of the PLK3 expressing cells are among the cells of cluster 5 (mostly old cells). Thus, PLK3 is upregulated in old cells.
Fig 8.
Gene ontology enrichment analysis of the five top-ranked genes.
The color gradients in A to C indicate the adjusted p-values. (A) The resulting Gene Ontology Biological Processes (GOBP). (B) The resulting GO Cellular Components (GOCC). (C) The resulting GO Molecular Functions (GOMF). (D) CNET plot visualizing the GOMFs that were associated with the five top-ranked genes.