Assignment #3

Assignment 3 - Data set Pathway and Network Analysis

---

Objective: Perform non-thresholded Gene set Enrichment Analysis and visualize the result using Cytoscape

Time estimated: 10 h; taken 20 h;

Date started: 2023-3-29; completed: 2023-4-4

---

Introduction

• As documented in Assignment #1, we used the normalized and mapped dataset with source:

  https://www-ncbi-nlm-nih-gov/geo/query/acc.cgi?acc=GSE104406	Aging Human Hematopoietic Stem Cells Manifest Profound Epigenetic Reprogramming of Enhancers That May Predispose to Leukemia (RNA-Seq of HSCe)

• To avoid repetitive calculation, the result of differential gene expression analysis was exported and imported using the following code:

  write.csv(output_hits, "HSCe_output_hits", row.names = TRUE)

  Then imported with read.table()

  Error: the legend of image wasn’t displayed properly Solution: add r fig MDS, fig.cap="\\label{fig:MDS}

Non-thresholded Gene set Enrichment Analysis

1. Data preperation

   • We calculathe the rank using: $Rank = -log_{10}(p-value) \cdot sign(logFC)$

2. Load geneset

   • We used genesets from the baderlab geneset collection containing GO biological process, no IEA and pathways. It is up-to-dated than GSEA default

3. GSEA

   • We used the default parameters, limit the geneset size within the range of 15 and 200 to ensure that the results has both specificity and generality, then run the test.

   

⚠️It is notable that 0 gene sets are significant at FDR < 25% for the enrichment for the phenotype group: aged, while 754 gene sets are significantly enriched at FDR < 25% for the young group.

• To resolve this issue, we redo the ranking procedure with logFC instead of modified rank, but this issue remains. We verified the biological background, as well as the technical properties of the RNA seq, but we cannot spot the issue because the ORA result is proper. Eventually the only possible explanation is that too few factor to be considered in the previously fitted model, for example gender or individuals in each group can also be crucial factors.

Visualize Analysis in Cytoscape

• We use the EnrichmentMap in Cytoscape to construct the graph, with parameters:
  • FDR q-value cutoff: 0.7 (otherwise will capture nothing)
  • Edge cutoff: 0.375
• We use AutoAnnotate to construct the annotation and the theme network, with parameters:
  • MCL clustering algorithm
  • Max word per label: 3
  • Min word occurrence: 1
  • adjacent word bonus: 8

Post Analysis

• We used the signature gene set of Transcription Factors in the newest from the Bader Lab geneset collection (inbuilt download), then we used a two-sided Mann_whitney testing with a threshold p value of 0.05.

  names	genes	largest_overlap	Mann_whitney
  SNRNP70	662	50	1.0545841977460668E-10
  CEBPZ	1184	37	3.3938411858613904E-8
  SALL4	1353	34	3.949491436006092E-6

• SALL4, CEBPZ, and SNRNP70 which are marked as related transcription factors with HSC or myeloid malignancies in previous researches

💡 Conclusion and outlook: GSEA encountered issue. Found evidences supporting that aging can affect the gene expression hence through the pathways impair functions.

---