Junction-Level Differential Gene Analysis (JDEG)

In this section, we perform junction-level differential gene analysis using the adjacency matrix. This analysis aims to identify genes that exhibit significant differences in junction-level expression between different conditions or cell types.

Step 1: Identify Junction Markers Using MAST

This step is identical to Step1of the EDEG analysis. The only difference is that the input_anndata file is changed to AdjacencyComp_PDAC.h5ad, which contains junction-level features instead of exon-level ones. The rest of the analysis pipeline remains the same, including the use of the MAST model through Seurat.

Step 2: Identify Junction-Level Markers Using Exon-Level Information

To determine gene-level differential expression, we now aggregate exon marker information by gene, enabling a higher-level view of differences.

The following function converts junction-level marker results from Seurat into gene-level statistical information. It uses the Stouffer method to combine junction p-values, and also computes average log2 fold changes.

from DOLPHIN.EDEG.generate_JDEG import run_jdeg

pd_JDEG = run_jdeg(seurat_output = "/mnt/data/kailu/00_scExon/10_GO_PDAC/02_model/02_exon_adj/MAST/AdjacencyComp_MAST_ductal.csv",
    output = "./PDAC_MAST_ductal_junction_final.csv")

pd_JDEG.head()

	Exon_names	p_val	avg_log2FC	pct.1	pct.2	p_val_adj	Gene_names	MAST_abs_avg_log2FC	MAST_stouffer_pval	MAST_stouffer_pval_adj_bonf
0	RPS26-12	4.578045e-194	-2.627707	0.949	0.989	6.177248e-189	RPS26	1.923962	6.474450e-153	6.062675e-149
1	FXYD2-22	5.593814e-170	-4.929407	0.299	0.994	7.547845e-165	FXYD2	4.119512	0.000000e+00	0.000000e+00
2	FXYD2-32	3.611924e-162	-4.767175	0.226	0.994	4.873642e-157	FXYD2	4.119512	0.000000e+00	0.000000e+00
3	RPL34-8	5.754681e-156	-2.005695	0.996	1.000	7.764906e-151	RPL34	2.007840	5.145729e-143	4.818461e-139
4	FXYD2-42	1.697355e-154	-4.520503	0.165	0.989	2.290275e-149	FXYD2	4.119512	0.000000e+00	0.000000e+00