Chapter 12 TPC Pipelines(2)

12.1 General steps

S1: Preset — Set personalized samples and tumor data with following general steps.

  • “Modify datasets”: View the dataset sources for each molecular type. When there are alternative datasets for the same molecular type, users can switch a suitable one based on purpose (referring to Chapter 10).
  • “Choose cancer”: Choose one cancer type for downstream association analysis.
  • “Filter samples”: Upon the selection of one tumor, further refine the samples based on tissue codes or personalized filtering module.
  • “Upload metadata”: Upload user-defined tumor data for joint analysis.
  • “Add signature”: Design custom molecular signature for joint analysis.

S2: Get data — Select and fetch identifier values; set grouping information.

S3: Analyze & Visualize — Modify diverse parameters; perform analysis; download results.

  • “Set analyzation parameters”: Select different statistical methods
  • “Set visualization parameters”: Adjust plot color, title text, etc.
  • “Download results”: Download three types of results, including raw data, detailed statistical results and plot results. (Note: No plot for batch-screen mode).

Notes: We will introduce various analysis scenarios of TPC using TCGA as example.

12.2 Correlation analysis

3 modes of correlation analysis

Figure 12.1: 3 modes of correlation analysis

12.2.1 Individual Mode

Perform the correlation analysis between any two identifiers of samples from one tumor.

The following is a simple example. (The steps not mentioned in above figure imply default operations.)

S1: Preset

  • “S1.2 Choose cancer”: choose the CHOL (Cholangiocarcinoma) cancer type
  • “S1.3 Filter samples”: choose the primary tumor samples via the quick filter.

S2: Get data

  • “S2.1 Get data for X-axis”: Molecular profile→mRNA Expression→TP53
  • “S2.2 Get data for Y-axis”: Infiltration→CIBERSORT→ Monocyte

S3: Analyze & Visualize

  • “S3.1 Set analysis parameters”: Spearman
  • “S3.2 Set visualization parameters”: Adjust the main title name.
Example of correlation analysis in individual mode

Figure 12.2: Example of correlation analysis in individual mode

12.2.2 Pan-cancer Mode

Perform the correlation analysis between any two identifiers of samples across multiple tumors.

The following is a simple example. (The steps not mentioned in above figure imply default operations.)

S1: Preset

  • “S1.1 Modify datasets”: Select the Norm_Count dataset for mRNA Expression dataset.
  • “S1.2 Choose cancer”: choose all TCGA cancer types.
  • “S1.3 Filter samples”: choose the non-normal samples via the quick filter.

S2: Get data

  • “S2.1 Get data for X-axis”: Molecular profile→mRNA Expression→HDAC1
  • “S2.2 Get data for Y-axis”: Molecular profile→mRNA Expression→APOE

S3: Analyze & Visualize

  • “S3.1 Set analysis parameters”: Pearson
  • “S3.2 Set visualization parameters”: Adjust the names of main title and axis title.
Example of correlation analysis in pan-cancer mode

Figure 12.3: Example of correlation analysis in pan-cancer mode

12.2.3 Batch Screen Mode

Perform the correlation analysis between multiple identifiers and one identifier of samples from one tumor.

The following is a simple example. (The steps not mentioned in above figure imply default operations.)

S1: Preset

  • “S1.2 Choose cancer”: choose the CHOL (Cholangiocarcinoma) cancer type
  • “S1.3 Filter samples”: choose the primary tumor samples via the quick filter.

S2: Get data

  • “S2.1 Get batch data for X-axis”: Molecular profile→mRNA Expression→ genes of HALLMARK_APOPTOSIS pathways
  • “S2.2 Get data for Y-axis”: Immune Infiltration→CIBERSORT→ Monocyte

S3: Analyze

  • “S3.1 Set analysis parameters”: Spearman
Example of correlation analysis in batch screen mode

Figure 12.4: Example of correlation analysis in batch screen mode

12.3 Comparison analysis

3 modes of comparison analysis

Figure 12.5: 3 modes of comparison analysis

12.3.1 Individual Mode

Perform the comparison analysis of one identifier between two groups of samples from one tumor.

The following is a simple example. (The steps not mentioned in above figure imply default operations.)

S1: Preset

  • “S1.2 Choose cancer”: choose the BRCA (Breast invasive carcinoma) cancer type
  • “S1.3 Filter samples”: choose the non-normal samples via the quick filter.

S2: Get data

  • “S2.1 Divide 2 groups by one condition”: Group1→Primary tumor samples; Group2→Metastatic tumor samples
  • “S2.2 Get data for comparison”: Molecular profile→mRNA Expression→POSTN

S3: Analyze & Visualize

  • “S3.1 Set analysis parameters”: Wilcoxon test
  • “S3.2 Set visualization parameters”: Remove x-axis title.
Example of comparison analysis in individual mode

Figure 12.6: Example of comparison analysis in individual mode

12.3.2 Pan-cancer Mode

Perform the comparison analysis of one identifier between two groups of samples across multiple tumors.

The following is a simple example. (The steps not mentioned in above figure imply default operations.)

S1: Preset

  • “S1.2 Choose cancer”: choose the all TCGA cancer types.
  • “S1.3 Filter samples”: choose the primary tumor samples via the quick filter. Further choose samples with age above 60 via the exact filter.

S2: Get data

  • “S2.1 Divide 2 groups by one condition”: Group1→ gene PTEN mutation; Group2→ gene PTEN wild-type.
  • “S2.2 Get data for comparison”: Tumor index → Tumor Stemness →RNAss

S3: Analyze & Visualize

  • “S3.1 Set analysis parameters”: t-test
  • “S3.2 Set visualization parameters”: adjust x-axis title.
Example of comparison analysis in pan-cancer mode

Figure 12.7: Example of comparison analysis in pan-cancer mode

12.3.3 Batch Screen Mode

Perform the comparison analysis of multiple identifier between two groups of samples from one tumor.

The following is a simple example. (The steps not mentioned in above figure imply default operations.)

S1: Preset

  • “S1.2 Choose cancer”: choose the BRCA (Breast invasive carcinoma) cancer type
  • “S1.3 Filter samples”: choose the non-normal samples via the quick filter.

S2: Get data

  • “S2.1 Divide 2 groups by one condition”: Group1→Primary tumor samples; Group2→Metastatic tumor samples
  • “S2.2 Get data for comparison”: Immune Infiltration→XCELL→ Monocyte

S3: Analyze

  • “S3.1 Set analysis parameters”: Wilcoxon test
Example of comparison analysis in batch screen mode

Figure 12.8: Example of comparison analysis in batch screen mode

12.4 Survival analysis

By default, both log-rank and cox regression survival analysis are performed for two groups of samples divided according to one condition. Here, we devised a switch “Whether use initial data before grouping?” at S3.1 steps with following function:

  • If its status is off , the above default analysis will performed;

  • If its status is on and the grouping condition is continuous:

    • For log-rank test, it will search the optimal cutoff with most significant survival difference;
    • For Cox regression, it will directly build the cox model based on the continuous variable.
3 modes of survival analysis

Figure 12.9: 3 modes of survival analysis

12.4.1 Individual Mode

Perform the survival analysis between two groups of samples generated by one identifier from one tumor.

The following is a simple example. (The steps not mentioned in above figure imply default operations.)

S1: Preset

  • “S1.1 Modify datasets”: Select the 27K Methylation datasets; Select the cg15206330 site to represent TP53 methylation
  • “S1.2 Choose cancer”: choose the BRCA (Breast invasive carcinoma) cancer type
  • “S1.3 Filter samples”: choose the non-normal samples via the quick filter.

S2: Get data

  • “S2.1 Select survival endpoint”: Select the OS event
  • “S2.2 Divide 2 groups by one condition”: Group1: TP53 methylation higher 50%, Group2: TP53 methylation lower 50%

S3: Analyze & Visualize

  • “S3.1 Set analysis parameters”: Log-rank text
  • “S3.2 Set visualization parameters”: Modify color; Set confidence interval.
Example of survival analysis in individual mode

Figure 12.10: Example of survival analysis in individual mode

12.4.2 Pan-cancer Mode

Perform the survival analysis between two groups of samples generated by one identifier across multiple tumor.

The following is a simple example. (The steps not mentioned in above figure imply default operations.)

S1: Preset

  • “S1.2 Choose cancer”: choose the BRCA cancer type
  • “S1.3 Filter samples”: choose the non-normal samples via the quick filter.

S2: Get data

  • “S2.1 Select survival endpoint”: Select the OS event
  • “S2.2 Divide 2 groups by one condition”: Group1: Signature value higher 10%, Group2: Signature value lower 10%

S3: Analyze & Visualize

  • “S3.1 Set analysis parameters”: Univariate Cox regression; Use the initial data.
  • “S3.2 Set visualization parameters”: Add main title.
Example of survival analysis in pan-cancer mode

Figure 12.11: Example of survival analysis in pan-cancer mode

12.4.3 Batch Screen Mode

Perform the survival analysis between two groups of samples generated by multiple identifiers from one tumor.

The following is a simple example. (The steps not mentioned in above figure imply default operations.)

S1: Preset

  • “S1.2 Choose cancer”: choose the all TCGA cancer types
  • “S1.3 Filter samples”: choose the primary tumor samples via the quick filter.
  • “S1.5 Add signature”: Design a molecular signature and add to custom metadata.

S2: Get data

  • “S2.1 Select survival endpoint”: Select the PFI event
  • “S2.2 Divide 2 groups by one condition”: Pathway activity→HALLMARK pathways→Group1: score higher 50%, Group2: score lower 50%

S3: Analyze

  • “S3.1 Set analysis parameters”: Log-rank test
Example of survival analysis in batch screen mode

Figure 12.12: Example of survival analysis in batch screen mode

12.5 Cross-Omics analysis

Recently, we have added one pipeline for TCGA cross-omics analysis to simultaneously explore the molecular features of multiple omics across pan-cancers.

12.5.1 Gene Cross-Omics Analysis

Perform the cross-omics analysis at the gene level, involving Gene expression, Gene Mutation, Gene CNV, Transcript expression, DNA methylation.

The following is a simple example. (The steps not mentioned in above figure imply default operations.)

S1: Preset

  • “S1.2 Choose cancer”: choose the all TCGA cancer types
  • “S1.3 Filter samples”: By default, select all normal samples and tumor samples.

S2: Get data

  • “S2.1 Select one gene”: Select gene TP53
  • “S2.2 Load mRNA/Mutation/CNV data”: Preload the mRNA/Mutation/CNV data of gene to save time in S3 step.
  • “S2.3 Load transcript data”: Select and load several transcript of gene, filter invalid transcripts with null value.
  • “S2.4 Load methylation data”: Select several CpG sits of gene and load methyaltion data.

S3: Analyze & Visualize : Analyze multi-omics data and visualize by funkyheatmap package.

Example of Gene Cross-Omics Analysis

Figure 12.13: Example of Gene Cross-Omics Analysis

Column-1: TCGA Names;

Column-2[Expression Profile]: 0-1 normalized median expression of gene in normal samples (include GTEx);

Column-3[Expression Profile]: 0-1 normalized median expression of gene in tumor samples;

Column-4[Expression Profile]: The significance symbol of gene comparison between tumor and normal samples via Wilcoxon test (P<0.001, “***”; P<0.01, “**”; P<0.05, “*”; P>0.05, “-”).

Column-5[Mutation Profile]: The distribution of gene mutation or wild status in tumor samples.

Column-6[Mutation Profile]: The percentage of gene mutation status in tumor samples.

Column-7[CNV Profile]: The distribution of gene copy number variation status in tumor samples (-2, homozygous deletion; -1, single copy deletion; 0, diploid normal copy; 1: low-level copy number amplification; 2: high-level copy number amplification).

Column-8[CNV Profile]: The percentage of gene copy number amplification (1,2) in tumor samples.

Column-9[CNV Profile]: The percentage of gene copy number deletion (-1, -2) in tumor samples.

Column-Selected transcript(s) of one gene[CNV Profile]: The median transcript expression in tumor samples.

Column-Selected CpG sites of one gene[Methylation Profile]: The median beta value of CpG sites in tumor samples.

12.5.2 Pathway Cross-Omics Analysis

Perform the cross-omics analysis at the pathway level, involving Gene expression, Gene Mutation, Gene CNV.

The following is a simple example. (The steps not mentioned in above figure imply default operations.)

S1: Preset

  • “S1.2 Choose cancer”: choose the all TCGA cancer types
  • “S1.3 Filter samples”: By default, select all normal samples and tumor samples.

S2: Get data

  • “S2.1 one pathway”: Select pathway HALLMARK_ADIPOGENESIS

S3: Analyze & Visualize : Analyze multi-omics data and visualize by funkyheatmap package.

Example of Pathway Cross-Omics Analysis

Figure 12.14: Example of Pathway Cross-Omics Analysis

Column-1: TCGA Names;

Column-2[Expression Profile]: 0-1 normalized median ssGSEA score of pathway in normal samples (include GTEx);

Column-3[Expression Profile]: 0-1 normalized median ssGSEA score of gene in tumor samples;

Column-4[Expression Profile]: The significance symbol of gene comparison between tumor and normal samples via Wilcoxon test (P<0.001, “***”; P<0.01, “**”; P<0.05, “*”; P>0.05, “-”).

Column-5[Mutation Profile]: The distribution of mutation or wild pathway gene status in tumor samples. If one of pathway genes is mutated, it is considered as mutation.

Column-6[Mutation Profile]: The percentage of pathway mutation status in tumor samples.

Column-7[Mutation Profile]: The mean counts of mutated pathway genes in tumor samples.

Column-8[CNV Profile]: The distribution of pathway gene copy number variation status in tumor samples. (Amp, homozygous deletion or single copy deletion; Non, diploid normal copy; Del, Copy number amplification.)

Column-9[CNV Profile]: The copy number amplification percentage of pathway genes in tumor samples.

Column-10[CNV Profile]: The copy number deletion percentage of pathway gene in tumor samples.