Differential Expression Analysis Pipeline

This guide provides an overview of what a Differential Expression Analysis report created by the BigMind team contains. For the integrated and automated analysis, we utilized functions from our custom R package, OmicsKit.

1. Preprocessing: Counts per Sample

Visualizing Samples Data

Sample metadata is crucial for connecting bioinformatics results to biological significance. The first step in our analysis is to clean the data and prepare it for downstream library requirements.

ID	Test	Condition	Group	Sex	Age	RIN
1	Undetected	Normal	A	F	65	6.0
2	Undetected	Normal	A	F	42	6.0
3	Undetected	Normal	A	M	63	5.5
4	Undetected	Normal	A	M	49	5.8
5	Undetected	Disease	B	M	51	5.2
6	Undetected	Disease	B	M	53	6.1
7	Detected	Normal	C	F	31	5.5
8	Detected	Normal	C	F	54	5.5
9	Detected	Normal	C	M	43	5.3
10	Detected	Normal	C	M	67	5.7
11	Detected	Normal	C	M	59	6.1
12	Detected	Normal	C	M	46	5.2
13	Detected	Normal	C	M	40	5.5
14	Detected	Disease	D	F	70	5.6
15	Detected	Disease	D	F	36	5.9
16	Detected	Disease	D	F	49	6.0
17	Detected	Disease	D	F	64	6.6
18	Detected	Disease	D	F	66	5.6
19	Detected	Disease	D	M	57	5.4
20	Detected	Disease	D	M	51	6.4
21	Detected	Disease	D	M	48	6.7

Managing Count Files

Identifying samples with abnormal counts can highlight outliers that may need to be excluded from analysis to avoid skewed results.

2. Exploring Confounders

Consider variables such as age, sex, batch effects, or other biological and technical factors that might impact gene expression.

Heatmap: Samples vs Samples

A heatmap is used to visualize the correlation between samples.

PCA: Principal Component Analysis

Principal Component Analysis (PCA) helps in understanding the variation between samples.

tSNE: t-Distributed Stochastic Neighbor Embedding

tSNE is another dimensionality reduction technique used to visualize high-dimensional data.

UMAP: Uniform Manifold Approximation and Projection

UMAP is used to visualize the structure of the data.

3. Differential Expression Analysis

DESeq2 is a widely used tool for analyzing differential gene expression in RNA-seq data. We performed our analysis by adjusting the analysis and improving filtering.

Display Results

baseMean	log2FoldChange	lfcSE	stat	pvalue	padj	ensembl	symbol	description	biotype
ENSG00000000003	4.8597885	-1.3065575	0.7629200	-1.7125747	0.0867908	NA	TSPAN6	tetraspanin 6 [Source:HGNC Symbol;Acc:HGNC:11858]	protein_coding
ENSG00000000005	0.0306208	0.5719487	4.3949775	0.1301369	0.8964581	NA	TNMD	tenomodulin [Source:HGNC Symbol;Acc:HGNC:17757]	protein_coding
ENSG00000000419	183.1677539	-0.0499737	0.1492696	-0.3347882	0.7377848	0.9888945	DPM1	dolichyl-phosphate mannosyltransferase subunit 1, catalytic [Source:HGNC Symbol;Acc:HGNC:3005]	protein_coding

Split Cases

When performing differential expression analysis of a study with 3 phenotypes, including the baseline, there are 10 mutually exclusive cases where genes can fall into.

Detectability

This function identifies genes with measurable expression levels across samples.

4. Plotting

Heatmap: Samples vs Genes

Heatmaps are used to visualize the expression levels of genes across different samples.

Volcano Plots

Volcano plots display statistical significance (p-value) versus magnitude of change (fold change).

Box-Scatter-Violin (BSV) Plots

BSV plots are used to visualize the distribution and expression levels of genes.

5. GO Enrichment

GO enrichment analysis with clusterProfiler helps interpret differential expression results by identifying biological pathways that are overrepresented in differentially expressed genes, providing insights into the underlying biological mechanisms.

Balloon Plots

Balloon plots visualize GO terms and their significance.

GO Terms Networks

GO terms networks show relationships between enriched GO terms.

6. Prognosis Model

Linear regression models are employed to analyze RNA-seq data and identify significant relationships between gene expression levels and clinical outcomes. These models help in developing predictive tools that can forecast disease progression or patient survival based on specific gene expression profiles. By leveraging these models, researchers can gain insights into the molecular underpinnings of disease and potential therapeutic targets.

ROC Curve

Receiver Operating Characteristic (ROC) curves are used to evaluate the performance of the prognosis models. The ROC curve plots the true positive rate (sensitivity) against the false positive rate (1-specificity) across different threshold settings. The area under the curve (AUC) provides a single measure of overall accuracy, with higher values indicating better predictive performance. These curves help assess the model's ability to distinguish between different clinical outcomes.