Fig. 4: Prioritization of LIME-identified genes for characterizing Parkinson disease cells.
From: Neural networks reveal novel gene signatures in Parkinson disease from single-nuclei transcriptomes
This figure uses astrocytes as a case study to demonstrate our approach for identifying the most biologically relevant genes identified by LIME. A Scatter plot showing the LIME feature importance Z-score for genes in the LIME optimal gene set for astrocytes. B Volcano plot showing the mean LIME feature importance Z-score across the Kamath et al., Wang et al., and Smajic et al. datasets and the gene expression log2 fold-change (Log2FC) between Parkinson disease (PD) and control astrocytes across all datasets. Only genes from the LIME-identified optimal gene set for astrocytes are included. C Heatmap showing the mean LIME feature importance Z-scores for genes in the astrocyte optimal gene set with an absolute Log2FC > 0.25. Grey tiles indicate genes that were not included in the cell type-specific optimal gene set. The genes retained for subsequent analysis are labelled. oligo oligodendrocytes, OPC oligodendrocyte precursor cells.
