IMPACT is a web-based platform for the exploration in IMmunotherapeutic Predictive And Cancer prognosTic biomarkers using genomic, transcriptomic and proteomic data from in-house, literature and public databases including The Cancer Genome Atlas (TCGA), Clinical Proteomic Tumor Analysis Consortium (CPTAC), International Cancer Genome Consortium (ICGC), Gene Expression Omnibus (GEO) and Singapore Oncology Data Portal(OncoSG).
ICIs Cohorts
Cancers
Samples
Parameters of cutpoint analysis:
Parameters of KM curve:
Parameters of Cox:
Parameters of Subgroup:
Parameters of Subgroup:
Parameters of GSVA:
Data Resources in IMPACT
IMPACT User Guide
IMPACT is a web-based platform for the exploration of IMmunotherapeutic Predictive And Cancer prognosTic biomarkers using genomic, transcriptomic and proteomics datasets from in-house, literature and public databases (TCGA, CPTAC, ICGC, GEO, OncoSG). IMPACT provides versatile functions for exploring predictive and prognostic biomarkers, including PredExplore, ProgExplore, Survival Analysis, Cox Regression, Subgroup Analysis, Cutpoint Analysis, Immunotherapy Response, Interaction Analysis, Immunogenicity Analysis, tumor microenvironment (TME) Correlation Analysis, TME Comparative Analysis, Mutation Profiles and User-defined function. Raw data and generated figures are available to research communities free of charge (accessibility of in-house data may need further requests). Different functions can be accessed via the menu bar at the top of the webpage.
Detailed function description and tutorial are as follows:
In the PredExplore module allows users to explore the value of genes of interest as immunotherapy biomarkers across multiple cohorts. The analysis can be performed in the following steps.
Step 1: Navigate to the PredExplore tab.
Step 2: Select one cancer type from the drop-down list (Figure 1A).
Step 3: Select one or more interested cohorts from the drop-down list (Figure 1B). By default, all cohorts were chosen.
Step 4: Select a predefined gene signature or pathway from the drop-down list, or input the gene symbol manually (Figure 1C). If using the predefined gene list, the genes in the pathway will be automatically filled into the input box. The users can generate a list up to 50 genes of interest.
Step 5: If two or more genes were input, the relationship among them needs to be considered (Figure 1D). “AND” means that samples will be stratified into mutated group based on all queried gene mutations. “OR” means that samples with as least one of the input gene mutations will be stratified into a mutated group.
When all parameters are defined, click “Submit” to start the analysis. The cohort will be stratified into the mutated and wild-type groups based on the queried genes. Then, the results of Cox regression and log-rank test, as well as differences in clinical factors between the two groups will be presented in a table (Figure 1E). Click the result can call out the Kaplan-Meier curves (PFS and/or OS), boxplots (TMB level, Neoantigen expression, Tumor purity, and PD-L1 expression), and forest plots of meta-analysis for PFS/OS (Figure 1F). The results and figures can be downloaded from “DLTable” and “DLGraph”, respectively.
Figure 1. The interface of Immunotherapy Predictive Biomarkers Exploration function.
ProgExplore: Cancer Prognostic Biomarkers Exploration
In the ProgExplore module, users can investigate the prognostic value of genes of interest across 40 cohorts of 19 cancer types. The analysis can be performed in the following steps.
Step 1: Navigate to the ProgExplore tab.
Step 2: Select a cancer type from the drop-down list (Figure 2A).
Step 3: Select a predefined gene signature or pathway from the drop-down list, or input the gene symbol manually (Figure 2B). If choosing a predefined gene list, the genes in the pathway will be automatically filled into the input box. Users can add more genes or delete the input gene symbols.The users can generate a list up to 10 genes of interest.
Step 4: If two or more genes are selected, the relationship among them needs to be considered (Figure 2C). “AND” means that samples will be stratified into mutated group based on all queried gene mutations. “OR” means that samples with as least one of the queried gene mutations will be stratified into a mutated group.
When all parameters are defined, click “Submit” to start the analysis. The cohort will be stratified into the mutated and wild-type groups based on the queried genes. Then, the results of Cox regression and log-rank test, as well as differences in clinical factors between the two groups will be presented in a table (Figure 2D). Click the result can call out the Kaplan-Meier curves (PFS and/or OS) and boxplots (TMB level, Neoantigen expression, Tumor purity, and PD-L1 expression) (Figure 2E). The results and figures can be downloaded from “DLTable” and “DLGraph”, respectively.
Figure 2. The interface of Cancer Prognostic Biomarker Exploration function.
Survival Analysis: Kaplan-Meier Curve
The Kaplan-Meier curve is used to estimate the prognostic value of mutation status or expression alteration of genes of interest. The analysis can be performed in the following steps.
Step 1: Navigate to the Survival Analysis tab and select the Kaplan-Meier curve subtab.
Step 2: Select the study type, cancer type and research dataset (Figure 3A).
Step 3: Select the data type, fill in the genes of interest, choose the mutation type or select the cutoff values for gene expression. If the mutation has been selected in Step 2, and two or more genes are filled, the relationship among them needs to be considered (Figure 3B).
Step 4: Select the biopsy time and the immune checkpoint inhibitor (ICI) type (Figure 3C).
Step 5: Select the line color and the position of the annotation text (Figure 3D).
Detailed results for the Kaplan-Meier curve plotting are listed in Figure 3E and 3F. Users can select the plots of different genes by clicking the tab on top of this panel. The “All Queried Genes” tab is to show the combined overall alteration results of the selected gene set. All the corresponding results and figures can be downloaded from “DLTable” and “DLGraph”, respectively.
Figure 3. The interface of survival analysis with Kaplan-Meier estimation.
Survival Analysis: Cox regression
Cox regression is used to evaluate the prognostic value of the genetic alterations/expression levels and clinical properties, including age, sex, stage, smoking, and TMB, et al. The analysis can be performed in the following steps.
Step 1: Navigate to the Survival Analysis tab and select the Cox regression subtab.
Step 2: Select the study type, cancer type and research dataset (Figure 4A).
Step 3: Select the data type, fill in the genes of interest, choose the mutation type or select the cutoff values for gene expression. If the mutation has been selected in Step 2, and two or more genes are filled, the relationship among them needs to be considered (Figure 4B).
Step 4: Select the clinical characteristics (Figure 4C).
Univariable and multivariable Cox regressions will be conducted using clinical properties with the mutation or expression information of selected genes. Forest plots and detailed results will be generated, as shown in Figure 4D and 4E, respectively. All the results and figures can also be downloaded by clicking from “DLTable” and “DLGraph”, respectively.
Figure 4. The interface of survival analysis using Cox Regression.
Survival Analysis: Subgroup Analysis
Subgroup analysis aims to investigate the interaction effects of mutation status or expression of genes of interest with related clinical characteristics. The analysis can be performed in the following steps.
Step 1: Navigate to the Survival Analysis tab and select the Subgroup Analysis subtab.
Step 2: Select the study type, cancer type and research dataset (Figure 5A).
Step 3: Select the data type, fill in the genes of interest, choose the mutation type or select the cutoff values for gene expression. If two or more genes have been filled, the relationship among them needs to be considered (Figure 5B).
Step 4: Select the interaction factor type (Figure 5C)
Users can check the forestplot of different genes by clicking the tabs of gene symbols as shown in Figure 5D. The corresponding data table to the forestplot is shown as in Figure 5E. Detailed result tables can be downloaded via the “DLTable”.
Figure 5. The interface for Subgroup Analysis.
Survival Analysis: Cutpoint Analysis
Cutpoint analysis aims to determine the optimal cutpoint for one or multiple continuous variables at once, using the Maximally Selected Rank Statistics method. The analysis can be performed in the following steps.
Step 1: Navigate to the Survival Analysis tab and select the Cutpoint Analysis subtab.
Step 2: Select the study type, data type and research dataset (Figure 6A).
Step 3: Select the dataset and fill in the genes or clinical characteristics of interest (Figure 6B).
Step 4: Select the line color (Figure 6C).
The line plot of HR and 95%CI of all proportions and detailed results will be generated, as shown in Figure 6D and 6E, respectively. All the results and figures can also be downloaded from “DLTable” and “DLGraph”, respectively.
Figure 6. The interface for Cutpoint Analysis function.
Survival Analysis: Immunotherapy Response
Immunotherapy response analysis is used to investigate differences in overall response rate (ORR) and clinical benefit between the genetic alterations/expression levels or clinical properties, including age, sex, stage, smoking, and TMB, et al. The analysis can be performed in the following steps.
Step 1: Navigate to the Survival Analysis tab and select the Immunotherapy Response subtab.
Step 2: Select the cancer type and research dataset (Figure 7A).
Step 3: Select the data type (Clinicopathologicals or Mutation) and the corresponding clinical factors or gene symbols and choose the mutation type of genes(Figure 7B).
Step 4: Select the biopsy time and the immune checkpoint inhibitors (ICIs) type (Figure 7C).
Step 5: Select the grouped color of the box plot (Figure 7D).
The stack plot, box plot and detailed results will be generated, as shown in Figure 7E and 7F, respectively. All the results and figures can also be downloaded from “DLTable” and “DLGraph”, respectively.
Figure 7. The interface for Immunotherapy Response function.
Interaction Analysis
Interaction Analysis is used to test interaction effects of gene mutation status (or expression levels) with treatment. The analysis can be performed in the following steps.
Step 1: Navigate to the Interaction Analysis tab.
Step 2: Select the dataset and cancer type (Figure 8A).
Step 3: Select and fill in the genes of interest and choose the mutation type. twoor more genes have been filled, the relationship among them needs to be considered (Figure 8B).
Step 4: Select the line color of different groups. (Figure 8C).
The Kaplan-Meier curves in two treatment groups and the detailed result tables are shown as in Figure 8D and 8E, respectively. The value of P for interaction is shown in the Kaplan-Meier curves. Users can also download the figures and results via “DLTable” and “DLGraph”, respectively.
Figure 8. The interface for Interaction Analysis function.
Immunogenicity Analysis
Immunogenicity analysis has included the continuous immune-related biomarkers and performed the comparative analysis of tumor purity, TMB, Neoantigen, and PD-L1 expression based on the mutation status or expression levels of selected biomarkers. The analysis can be performed in the following steps.
Step 1: Navigate to the Immunogenicity Analysis tab.
Step 2: Select the study type, cancer type and dataset (Figure 9A).
Step 3: Select X-axis variables: the data type, the genes of interest and the mutation type or the cutoff values for gene expression. If the mutation has been selected in step 2, and two or more genes are filled, the relationship among them should be considered (Figure 9B).
Step 4: Select Y-axis variables, including Clinicopathologicals, mRNA expression and ssGSEA score of predefined pathways or gene signatures (Figure 9C).
Step 5: Select the color of the box plot or scatter plot of two groups (Figure 9D).
The box plots or scatter plots and detailed results tables will be generated as shown in Figure 9E and 9F, respectively. The results tables and figures can also be downloaded from “DLTable” and “DLGraph”, respectively.
Figure 9. The interface of Immunogenicity Analysis function.
TME Analysis: TME Correlation Analysis
In the Tumor microenvironment (TME) Correlation Analysis, users can investigate the correlations of selected genes with selected immunomodulators or signatures, including gene sets (Chemokine, Receptor, MHC, Immunoinhibitor, Immunostimulator, et al.) and pathways (Immune-related signatures, Key signatures, Oncogenic Signalling, et al.). The analysis can be preformed in following steps.
Step 1: Navigate to the TME Analysis tab and select the TME Correlation Analysis subtab.
Step 2: Select the study type, cancer type and dataset (Figure 10A).
Step 3: Select and fill in the genes of interest (Figure 10B).
Step 4: Select the immune genes or signatures (Figure 10C).
Step 5: Select the method used for correaltionship (Figure 10D).
Step 6: Select the figure type and graphic color (Figure 10E).
The heatmap of the correlation analysis can be found in Figure 10F and the detailed results can be found in Figure 10G. The results and figures can be downloaded from “DLTable” and “DLGraph”, respectively.
Figure 10. The interface of TME Correlation Analysis function.
Tumor microenvironment: TME Comparative Analysis
In the TME Comparative Analysis module, users can investigate the immune-related gene or signature differences in low- and high-expressed groups of selected genes. The analysis can be performed in the following steps.
Step 1: Navigate to the TME Analysis tab and select the TME Comparative Analysis subtab.
Step 2: Select the study type, cancer type and dataset (Figure 11A).
Step 3: Select and fill in the genes of interest and the cutoff values for gene expression (Figure 11B).
Step 4: Select the immune-related genes or pathways (Figure 11C). If the pathway has been selected, the ssGSEA score will be calculated.
Step 5: Select the graphic color of low and high-group or expression(Figure 11D).
The heatmap of the gene expression or ssGSEA score of pathways can be found in Figure 11E and the difference between low- and high-group can be found in Figure 11F. The results and figures can be downloaded from “DLTable” and “DLGraph”, respectively.
Figure 11. The interface of the TME Comparative Analysis function.
Mutation Profiles
The tumor Mutation Profile function helps visualize the mutation status of selected genes in groups stratified by the mutation status or expression levels of genes of interest. The analysis can be performed in the following steps.
Step 1: Navigate to the Mutation Profiles tab.
Step 2: Select the study type, cancer type and dataset (Figure 12A).
Step 3: Select the data type and fill in the genes of interest and the mutation type or the cutoff values for gene expression. If the mutation has been selected in Step 2, and two or more genes are filled, the relationship among them needs to be considered (Figure 12B).
Step 4: Select and enter the displayed gene symbols (Figure 12C).
Step 5: Select the mutation type of genes (Figure 12D).
Step 6: Exclude known frequently mutated genes (FLAGs) (Figure 12E).
Step 7: Customized the colors of different groups (Figure 12F).
Results are shown in Figures 12G and 12H. The results and figures can be downloaded from “DLTable” and “DLGraph”, respectively.
Figure 12. The interface of Tumor Mutation Profiles visualization.
Data availability
All the datasets used in IMPACT are listed as a table on the “Data” page. Detailed information on each dataset can be found in the table as shown in Figure 13. We provide the download links for all the publicly available datasets. Further requests will be needed to download our in-house datasets.
Figure 13. The screenshot of dataset information used in IMPACT.
Introduction
Identifying suitable biomarkers for immune checkpoint inhibitors (ICIs) could efficiently screen beneficiaries and improve the clinical use of ICIs currently. However, the biomarker discovery process could be largely accelerated with an integrated analytical platform combining large-scale high-quality datasets and a comprehensive solution. Thus, we developed a publicly available web server, IMPACT, based on both public and in-house datasets, which can facilitate the comprehensive investigation and visualization of predictive or prognostic biomarkers, relevant interaction effects, and biological mechanisms.
Up to now, IMPACT contains 6,276 patients treated with ICIs curated from 24 public datasets and 3 in-house cohorts, allows users to upload their own data, and consists of 11 function modules:
All the analyses are served to solve the following three clinical challenges:
Understanding ICIs biomarkers with synergistic or interaction effects;
Identifying a predictive biomarker specified for ICIs;
Exploring the functional mechanism of the biomarkers of interest;
Compared with other available web-based platforms, IMPACT exclusively provides interaction effect analyses module and it also optimized several conventional functions for discovering novel biomarkers, including but not limited to , selecting specific variant types, automatically screening co-mutations among multiple genes, and exploring cut-off values for gene expression biomarkers, all of which have been demonstrated with examples in the manuscript.
We give our special thanks to Shiny Gallery for sharing the excellent template. R packages of surviminer were adopted to implement IMPACT. We acknowledge their contributions to our work.
Contact us
If you have any questions about the IMPACT, please feel free to contact: Zhijie Wang (jie_969@163.com), Yutao Liu (liuyutao2529@126.com) and Wenchuan Xie (wenchuan.xie@gmail.com). When describing the issues, please attach the related screenshots and provide the information about the problem as detailed as possible, including the information of the browser you used.
As a new platform, we indeed have some plans and ongoing projects to empower it continuously:
Regular Data Resource Updates: We will update the data resources every six months to ensure the timely inclusion of new datasets.
Addition of Methylation Data: We will incorporate methylation data of immunotherapy to expand the available biomarker data in IMPACT. Now, we are collecting relevant data
Feature Updates Based on User Feedback: We will actively gather and incorporate user feedback to update and enhance the functionality of IMPACT.
