FOR RESEARCH USE ONLY. Not for use in diagnostic procedures.

IO 360 Data Analysis Report

Summary

    • Customer: Patricia Smith
    • Institute: Cancer Institute
    • Number of Samples: 144
    • Site of nCounter Run: NanoString Technologies
    • 360 Analyst: Nanostring Scientist
    • Scientific Reviewer: Nanostring IO Lead
    • Date: October 09, 2020

    This is a demo report used to showcase the functionality of the IO360 Standard Time Series Report.

Differential Expression Analysis

Response

Mutational Status

Survival Analysis

Progression Free Survival

Loss Signatures

Time-Series Analysis

Time Series Analysis

Quality Control

QC Summary for Response Grouping Variable

QC Summary for Mutational Status Grouping Variable

QC Summary for Progression Free Survival

Heatmap Overview

Heatmaps display gene expression values (or signature scores) across samples. These values are generally centered and scaled within each gene or signature for better graphical representation. Each tile within the heatmap displays a color that indicates the relative value of gene expression or signature score for the corresponding sample. If the samples can be sorted into groups, and those groups are provided to the analyst, they are indicated by the colored tiles across the top of the heatmap and described in the key to the right. The samples are sorted in order to place tiles with similar expression near each other to more easily identify patterns. The levels of relatedness are organized by unsupervised hierarchical clustering.

All Signatures

The 'All Signatures' heatmap uses unsupervised hierarchical clustering to show relatedness among signature scores for each sample. Scores are scaled by signature to have a mean of zero and a standard deviation of one. The standardized signature scores are truncated at ± 3 standard deviations to preserve greater clarity in color change within the largest proportion of data (99% of the data should fall within ± 3 standard deviations of the mean). Sample annotations are listed at the top of the heatmap. The signatures are displayed in rows and listed to the right of the heatmap. Each column is a unique sample, with a sample label displayed below the heatmap if there are 36 or fewer samples in the analysis.

All Genes

The 'All Genes' heatmap uses unsupervised hierarchical clustering to group normalized gene expression by sample. Expression values are scaled by gene to have a mean of zero and a standard deviation of one and then truncated at ± 3 standard deviations to preserve greater clarity in color change within the largest proportion of data (99% of the data should fall within ± 3 standard deviations of the mean). Sample annotations are listed at the top of the heatmap. The genes are displayed in rows. Each column is a unique sample, with a sample label displayed below the heatmap if there are 36 or fewer samples in the analysis.

Genes Within Signatures

The 'Genes Within a Signature' heatmap uses unsupervised hierarchical clustering to group normalized gene expression for genes within the signature. The different signatures are chosen from the dropdown menu. Scores are scaled by gene to have a mean of zero and a standard deviation of one and then truncated at ± 3 standard deviations to preserve greater clarity in color change within the largest proportion of data (99% of the data should fall within ± 3 standard deviations of the mean). Sample annotations are listed at the top of the heatmap and included genes are listed to the right of the heatmap. Only signatures comprised of 5 or more genes are present in this section.

Response

Differential Expression - Response Analysis

Differential expression analysis evaluates how different two groups are based on their gene/signature expression profiles. In Response Analysis, differential response to therapy, disease progression state, or binary category is used to classify each group, as defined for each sample in the annotation file. Comparison of the mean and range of expression between groups is used to understand if there are statistical similarities or differences.

The first plot on the page is a volcano plot that displays the fold-change and significance of the difference in signature scores between the two groups. Below the volcano plot, a forest plot sorts and summarizes the data for all the signatures. Next to the forest plot, a box plot displays the individual scores of samples in each group for a given signature that can be selected using the drop-down menu. Below the forest and box plot, a table is provided for the signature data. Beneath the signature section, a volcano plot and data table are provided to display the gene level analysis.

Following the Differential Expression Analysis, a Receiver Operator Characteristic (ROC) plot is displayed and presents the capacity of each signature to predict response. This plot may be stratified by a variable to investigate possible differences between the ROC curves for the different stratums.

All Signatures

Volcano Plot

The 'All Signatures' volcano plot displays each signature's difference between the response category, represented along the x-axis, with the significance (p-value) along the y-axis. Signatures that have greater statistical significance will produce points that are both larger and darker in hue, in addition to appearing higher on the plot. Signatures that have greater differential expression versus the baseline group appear further from the center of the plot. Signatures further to the right indicate an increase in expression and signatures further to the left indicate a decrease in expression relative to the baseline group. Horizontal lines indicate 0.01 and 0.05 adjusted p-values; when the adjusted p-values range higher than 0.05, the thresholds are not shown in the plot.

Forest and Box Plots

The 'All Signatures' forest plot shows the differential expression means and 95% confidence intervals between response variables, for each signature on an unadjusted scale. The vertical axis is shown at fold change equal to zero, indicating equivalent expression between the groups. As the marker shifts from the center line there is an increase (shift to the right), or decrease (shift to the left), in the differential expression of that signature when compared to the baseline group (represented as the vertical line at zero). The shape of the marker in each box indicates whether there is a significant difference in the signature as assessed by univariate analysis (note that this significance is not adjusted for multiple comparisons). A signature is considered significant if the 95% confidence interval (the horizontal line of the signature) does not cross the vertical axis representing the baseline group and therefore no difference to that baseline group, again not adjusted for multiple comparisons. If there is a complete absence of significant findings, no legend will accompany the plot.

Test Results

The 'All Signatures' results table provides the fold-change values for each signature used to generate the volcano plot and forest plot. This table reports the signature name, variable, group/response, Log(2) transformed fold change, 95% lower confidence of the mean limit, 95% upper confidence of the mean limit, students t-test distribution score, unadjusted significance (p-value), and significance adjusted for multiple tests (FDR).

All Genes

Volcano Plot

The 'All Genes' volcano plot displays each gene's fold change (or difference on the Log(2) scale) and significance (p-value). Genes that have greater statistical significance appear higher on the plot with larger, darker points, while genes that have greater differential expression appear further from the center of the plot. Genes further to the right indicate an increase in expression and genes further to the left indicate a decrease in expression relative to the baseline group. Horizontal lines indicate 0.01 and 0.05 adjusted p-values; when the adjusted p-values range higher than 0.05, the thresholds are not shown in the plot.

Test Results

The 'All Genes' results table provides the fold-change values for each gene used to generate the 'All Genes' volcano plot. This table reports the gene name, variable, group/response, Log(2) transformed fold-change, 95% lower confidence of the mean limit, 95% upper confidence of the mean limit, students t-test distribution score, unadjusted significance (p-value), and significance adjusted for multiple tests (FDR).

Receiver Operator Characteristic

All Signatures

Receiver Operator Characteristic (ROC) plots display the capacity of each signature to predict response. This plot may be stratified by a variable to investigate possible differences between the ROC curves for the different stratums. ROC plots are generated for the signature chosen from the dropdown menu and stratified by dichotomizing the signature using its median signature score as a threshold. The true positive rate (sensitivity) and false positive rate (1-specificity) are graphed and used to calculate the Area Under the Curve (AUC); a metric of predictive power. Signatures that are more predictive of response, have curves bent toward the upper left corner of the plot. This yields an AUC that approaches an upper bound of 1.0 at which point prediction is perfect. Signatures that are not predictive appear as diagonal lines from the lower left to the upper right corner of the plot. These curves have an AUC of approximately 0.5. The AUC value for each group in the analysis is displayed to the right of the plot.

Mutational Status

Differential Expression - Group Analysis

Differential expression analysis evaluates whether we can detect two or more groups as different based on their gene/signature expression profiles. In Grouping Analysis, intrinsic differences in the samples such as patient characteristics, or treatment arms of an investigational study, are used to define each group as specified in the annotation file. Comparison of the mean and range of expression between groups is used to understand if there are statistical differences.

The first plot on the page is a volcano plot that displays the fold-change and significance of difference between groups for the signature scores. Below the volcano plot, a forest plot summarizes mean differences between groups with their confidence interval for all the signatures. Next to the forest plot, a box plot displays the individual scores of samples in each group for a given signature that can be selected using the drop-down menu. Below the forest and box plot, a table is provided for the signature data. Beneath the signature section, a volcano plot and data table are provided to display the gene level analysis.

All Signatures

Volcano Plot

The 'All Signatures' volcano plot displays each signature's difference between the response category, represented along the x-axis, with the significance (p-value) along the y-axis. Signatures that have greater statistical significance will produce points that are both larger and darker in hue, in addition to appearing higher on the plot. Signatures that have greater differential expression versus the baseline group appear further from the center of the plot. Signatures further to the right indicate an increase in expression and signatures further to the left indicate a decrease in expression relative to the baseline group. Horizontal lines indicate 0.01 and 0.05 adjusted p-values; when the adjusted p-values range higher than 0.05, the thresholds are not shown in the plot.

Forest and Box Plots

The 'All Signatures' forest plot shows the differential expression means and 95% confidence intervals between response variables, for each signature on an unadjusted scale. The vertical axis is shown at fold change equal to zero, indicating equivalent expression between the groups. As the marker shifts from the center line there is an increase (shift to the right), or decrease (shift to the left), in the differential expression of that signature when compared to the baseline group (represented as the vertical line at zero). The shape of the marker in each box indicates whether there is a significant difference in the signature as assessed by univariate analysis (note that this significance is not adjusted for multiple comparisons). A signature is considered significant if the 95% confidence interval (the horizontal line of the signature) does not cross the vertical axis representing the baseline group and therefore no difference to that baseline group, again not adjusted for multiple comparisons. If there is a complete absence of significant findings, no legend will accompany the plot.

Test Results

The 'All Signatures' results table provides the fold-change values for each signature used to generate the volcano plot and forest plot. This table reports the signature name, variable, group/response, Log(2) transformed fold change, 95% lower confidence of the mean limit, 95% upper confidence of the mean limit, students t-test distribution score, unadjusted significance (p-value), and significance adjusted for multiple tests (FDR).

All Genes

Volcano Plot

The 'All Genes' volcano plot displays each gene's fold change (or difference on the Log(2) scale) and significance (p-value). Genes that have greater statistical significance appear higher on the plot with larger, darker points, while genes that have greater differential expression appear further from the center of the plot. Genes further to the right indicate an increase in expression and genes further to the left indicate a decrease in expression relative to the baseline group. Horizontal lines indicate 0.01 and 0.05 adjusted p-values; when the adjusted p-values range higher than 0.05, the thresholds are not shown in the plot.

Test Results

The 'All Genes' results table provides the fold-change values for each gene used to generate the 'All Genes' volcano plot. This table reports the gene name, variable, group/response, Log(2) transformed fold-change, 95% lower confidence of the mean limit, 95% upper confidence of the mean limit, students t-test distribution score, unadjusted significance (p-value), and significance adjusted for multiple tests (FDR).

Progression Free Survival

Survival Analysis - Progression-Free

Survival analysis evaluates whether the change in disease state over time, as observed in two or more groups, are associated with an expression profile or gene signature. For Progression-Free Survival Analysis, the survival of a patient is observed over time until a censoring event such as patient departure from the study, disease recovery, or death occurs. Signature score expression is stratified amongst each sample grouping to determine if there is a positive or negative relationship between a given signature and a change in disease state of the patient.

The first plot on the page is a volcano plot that displays the hazard ratio and significance (p-value) both estimated from a univariate cox regression model. Below the volcano plot, a forest plot displays the hazard ratio based upon the overall survival difference based upon a signature's expression. Following the forest plot is a table that displays the expression data and hazard ratios used to generate the forest plot. Below the forest plot, a Kaplan-Meier curve for each signature is rendered. For every Kaplan-Meier curve, a risk table displays the number of patients remaining at each observation time point. Below the Kaplan-Meier curve, a survival table provides all data points in above analysis.

Hazard Ratio

Volcano Plot

The survival volcano plot displays each signature's hazard ratio and significance (p-value). Signatures that have greater statistical significance appear higher on the plot with darker, larger points, while signatures that have more extreme hazard ratios appear further from the center of the plot. Signatures further to the right are associated with decreased risk of an event relative to the baseline and signatures further to the left are associated with greater risk of an event relative to the baseline. Horizontal lines indicate 0.01 and 0.05 adjusted p-value; when the adjusted p-values range higher than 0.05, the thresholds are not shown in the plot.

Forest Plot

The survival forest plot shows the distribution of log-hazard ratio median and unadjusted 95% confidence intervals for each signature across all samples. Confidence intervals for the median hazard are depicted along the x-axis, as 95% confidence intervals. Each signature name is listed on the y-axis and is sorted by the median value from highest to lowest. A vertical axis is shown at a ratio value equal to one (log-hazard of zero) which indicates no difference from the baseline group. If there is a complete absence of significant findings, no legend will accompany the plot.

Model Table

The Model Table provides the hazard ratio values for each signature used to generate the forest plot. This table reports the signature name, log-hazard ratio (coef), 95% lower confidence of the mean limit, 95% upper confidence of the mean limit, raw hazard ratio (exp(coef)), hazard ratio standard error (coef), z score, unadjusted significance (p-value), and significance adjusted for multiple tests (FDR).

Survival Probabilities

Survival Curves

The Survival Plot is a Kaplan-Meier curve that shows the association between a signature score and survival. Samples are categorized into groups based on the distribution of expression of a signature score within the cohort, then the survival probability of each group is presented over time.

Survival Table

The Survival Table contains all the data used to generate the Kaplan-Meier plot and Risk Table. For each signature, and its expression level sub-strata, the table displays the time of an event, the event type (n.event or n.censor), the number of patients at risk (n.risk), strata hazard ratio (surv), standard error, 95% lower confidence of the mean hazard ratio, and 95% upper confidence of the mean hazard ratio.

Single Sample Analysis

The single-sample analyses present a way of looking at the expression data that focuses on the expression levels within an individual sample, rather than groups or the entire cohort. Within this tab, the wheel plots enable a summary of the overall expression for all signatures within a sample, as comparisons of it's expression to the expression of all the samples provided. The scatter plots show the sample relative to the expression of all the samples in the report. Each scatter plot shows the signatures versus the TIS signature. The expression of the individual samples is highlighted as a larger dot within the expression of all the samples. The signature scores table provides the numerical score of each signature for each individual sample.

Wheel Plots

The Wheel Plot depicts the relative expression of each signature for an individual sample. Signatures are grouped based on the biological process in which they belong, and the TIS score is shown as a radial arc around its numeric score rounded to two significant digits. Signature scores are represented as radial projections, with negative score values highlighted by a grey outline. An individual sample can be selected from the drop-down menu.

Scatter Plots

The Scatter Plots depict the association of TIS to other IO 360 signatures. An individual sample can be selected from the drop-down menu and is represented on the graphs as a larger point.

Signature Scores

The signature scores for each sample are listed in this table. Typically, the signatures are computed as a weighted linear combination of Log(2) gene expression values. The weights applied sum to one. Thus, each unit increase in score corresponds to a doubling of the biological processes that are measured. A notable exception to this general method are the calculations for TIS and loss signatures; these are described in more detail on the Methods page.

Time Series Analysis

Changes to gene expression patterns and signature scores over time reflect a changing tumor microenvironment and may be useful for understanding the impact of therapy or natural evolution of the cancer. By comparing signature scores at later time points to the score at baseline, changes in the tumor-immune landscape can be visualized.

At the top of the page, each of the signature scores is presented in a spaghetti plot that depicts the average signature score for each level of a grouping variable at each time point. Below that, a separate spaghetti plot displays individual sample scores for a given signature; the signature displayed can be selected using the associated drop-down menu.

Spaghetti Plots - By Signature

The signature spaghetti plots display the mean signature score for each level of a grouping variable at each time point. The signature scores are represented on the y-axis, and the x-axis corresponds to time points represented as categories (i.e. the axis is not scaled to the numeric value of the time points). Connecting lines between each point visualize the change in signature score over time.

The 'By Signature' results table displays the mean score and standard error for each signature used to generate the above spaghetti plots. This table reports the signature name, time point, secondary grouping variable (when applicable), mean score, and standard error. Mean score and standard error are indicated as NA for single sample groupings.

Spaghetti Plots - By Sample

The sample spaghetti plots section displays the individual sample scores for a single signature. The signature data displayed can be selected using the dropdown menu. This plot visualizes the heterogeneity of scores within a group and can help to identify outlier samples.
The 'By Sample' results table provides the signature score data for each sample analyzed, indicating the time point and grouping variable assigned to each.

Differential Expression

Differential expression analysis evaluates whether we can detect two or more groups as different based on their gene/signature expression profiles. On this page, comparisons between the baseline time point and each other time point, as specified in the annotation file, are used to define each group. Comparison of the mean and range of expression between groups is used to understand if there are statistical differences.

The first plot on the page is a volcano plot that displays the fold-change and significance of difference between groups for the signature scores. Below the volcano plot, a forest plot summarizes mean difference between groups with their confidence interval for all signatures. Next to the forest plot, a box plot displays the individual scores of samples in each group for a given signature that can be selected using the dropdown menu. Below the forest and box plot, a table is provided for the signature data. Beneath the signature section, a volcano plot and data table are provided to display the gene level analysis.

All Signatures

Volcano Plots

The 'All Signatures' volcano plot displays each signature’s difference between the baseline and the subsequent time points. Time point comparisons can be selected via the dropdown menu. Each signature’s difference between the time points is represented along the x-axis, with the significance (p-value) along the y-axis. Signatures that have greater statistical significance will produce points that are both larger and darker in hue, in addition to appearing higher on the plot. Signatures that have greater differential expression versus the baseline group appear further from the center of the plot. Signatures further to the right indicate an increase in expression and signatures further to the left indicate a decrease in expression relative to the baseline group. Horizontal lines indicate 0.01 and 0.05 adjusted p-values; when the adjusted p-values range higher than 0.05, the thresholds are not shown in the plot.

Forest and Box Plots

The 'All Signatures' forest plot shows the differential expression means and 95% confidence intervals between time points for each level of a secondary grouping variable, for each signature on an unadjusted scale. The vertical axis is shown at fold change equal to zero, indicating equivalent expression between the groups. As the marker shifts from the center line there is an increase (shift to the right), or decrease (shift to the left), in the differential expression of that signature when compared to the baseline group (represented as the vertical line at zero). The shape of the marker in each box indicates whether there is a significant difference in the signature as assessed by univariate analysis (note that this significance is not adjusted for multiple comparisons). A signature is considered significant if the 95% confidence interval (the horizontal line of the signature) does not cross the vertical axis representing the baseline group and therefore no difference to that baseline group, again not adjusted for multiple comparisons. If there is a complete absence of significant findings, no legend will accompany the plot.

The accompanying box plot displays the signature scores for individual samples at each time point, along with the median, upper and lower quartiles, and 95% confidence interval.
The 'All Signatures' results table provides the fold-change values for each signature used to generate the volcano plot and forest plot. This table reports the signature name, variable (when applicable), time point comparison, Log(2) transformed fold change, 95% lower confidence of the mean limit, 95% upper confidence of the mean limit, students t-test distribution score, unadjusted significance (p-value), and significance adjusted for multiple tests (FDR).

All Genes

Volcano Plots

The 'All Genes' volcano plot displays each gene's fold change (or difference on the Log(2) scale) and significance (p-value) for comparison between time points. Genes that have greater statistical significance appear higher on the plot with larger, darker points, while genes that have greater differential expression appear further from the center of the plot. Genes further to the right indicate an increase in expression and genes further to the left indicate a decrease in expression relative to the baseline group. Horizontal lines indicate 0.01 and 0.05 adjusted p-value thresholds.
The 'All Genes' results table provides the fold-change values for each gene used to generate the 'All Genes Time Series' volcano plot. This table reports the gene name, variable (when applicable), time point comparison, Log(2) transformed fold-change, 95% lower confidence of the mean limit, 95% upper confidence of the mean limit, students t-test distribution score, unadjusted significance (p-value), and significance adjusted for multiple tests (FDR).

Loss Signature

Unlike the other signatures in the IO 360 report, which are weighted linear sums of gene expression, the loss signatures measure the decreased expression of a gene within a pathway where genes are typically expressed at constant rations. This is biologically relevant when the loss of expression of one gene in a pathway causes disfunction in that pathway. The exception is the signature for Hypermutation, which is a weighted linear gene signature, but it is used in the calculation of the MSI predictor and so is included.

Each signature may be viewed as a waterfall plot using the drop-down menu. Each sample is displayed as an individual bar in the plot. Annotations with each sample are represented as colored tiles below the graph. Below the colored tiles, the individual genes in each signature are displayed as box plots. Underneath the gene level box plots, the five loss signatures are displayed as box plots with individual samples colored by the grouping variables. Finally, a table with the numerical values for each score is located at the bottom of the page.

Loss Signatures - Time 1

Loss Signature Waterfall Plots

The Loss Signature waterfall plot displays the selected loss signature for all samples. For AMP Loss, JAK-STAT Loss and MMR Loss, the line at which loss is significant falls at zero. These plots are re-scaled so that the scores are deviations from the threshold over which a loss of function is defined. The scores are then reversed for lower values to be depicted as a loss. Bars that fall below the zero line indicate a potential loss for that sample. Bars above the threshold do not indicate a loss. There is a potential for some samples to show an indication that loss may be starting. These samples fall under the borderline threshold which is drawn at 1.

The hypermutation graph displays the hypermutation scores scaled to have a mean of zero and standard deviation of one and then outliers are truncated at ± 3 standard deviations. The MSI predictor is a combined signature from the MMR Loss signature and the hypermutation signature.

Loss Signature Box Plots

This set of box plots presents each loss signature within its own pane and colored by the levels for the group selected. A selector widget allows toggling between any group variables requested.

Loss Signatures - Time 2

Loss Signature Waterfall Plots

The Loss Signature waterfall plot displays the selected loss signature for all samples. For AMP Loss, JAK-STAT Loss and MMR Loss, the line at which loss is significant falls at zero. These plots are re-scaled so that the scores are deviations from the threshold over which a loss of function is defined. The scores are then reversed for lower values to be depicted as a loss. Bars that fall below the zero line indicate a potential loss for that sample. Bars above the threshold do not indicate a loss. There is a potential for some samples to show an indication that loss may be starting. These samples fall under the borderline threshold which is drawn at 1.

The hypermutation graph displays the hypermutation scores scaled to have a mean of zero and standard deviation of one and then outliers are truncated at ± 3 standard deviations. The MSI predictor is a combined signature from the MMR Loss signature and the hypermutation signature.

Loss Signature Box Plots

This set of box plots presents each loss signature within its own pane and colored by the levels for the group selected. A selector widget allows toggling between any group variables requested.

Loss Signatures - Time 3

Loss Signature Waterfall Plots

The Loss Signature waterfall plot displays the selected loss signature for all samples. For AMP Loss, JAK-STAT Loss and MMR Loss, the line at which loss is significant falls at zero. These plots are re-scaled so that the scores are deviations from the threshold over which a loss of function is defined. The scores are then reversed for lower values to be depicted as a loss. Bars that fall below the zero line indicate a potential loss for that sample. Bars above the threshold do not indicate a loss. There is a potential for some samples to show an indication that loss may be starting. These samples fall under the borderline threshold which is drawn at 1.

The hypermutation graph displays the hypermutation scores scaled to have a mean of zero and standard deviation of one and then outliers are truncated at ± 3 standard deviations. The MSI predictor is a combined signature from the MMR Loss signature and the hypermutation signature.

Loss Signature Box Plots

This set of box plots presents each loss signature within its own pane and colored by the levels for the group selected. A selector widget allows toggling between any group variables requested.

Loss Signature Scores

Quality Control Details

Summary of QC Results

This page provides a summary of the quality control metrics used to assess the technical performance of the nCounter profiling assay in this study. First, housekeeper genes assess sample integrity by comparing the observed value versus a predetermined threshold for suitability for data analysis. The machine performance is assessed using percentage of fields of view that were attempted versus those successfully analyzed. The binding density of the probes within the imaging area, ERCC linearity, and limit of detection are used as readouts of the efficiency and specificity of the chemistry of the assay. Any sample deemed as failing any one of these QC checkpoints will be removed from the analysis.

Housekeeping Genes QC

This plot shows the geometric mean of housekeeper genes in each sample. Samples with low housekeeper signal suffer from either low sample input or low reaction efficiency. Ideally the geometric mean of counts will be above 100 for all samples, and a minimum geometric mean of 32 counts is required for analysis. Samples in-between these two thresholds are considered in the analysis, but results from these "borderline" samples should be treated with caution.

Imaging QC

This metric reports the percentage of fields of view (FOVs) the Digital Analyzer or SPRINT was able to capture. At least 75% of FOVs should be successful to obtain robust data.

Binding Density QC

The binding density represents the concentration of barcodes measured by the instrument in barcodes per square micron. The Digital Analyzer may not be able to distinguish each probe from the others if too many are present. The ideal range for assays run on an nCounter MAX or FLEX system is 0.1 - 2.25 spots per square micron and assays run on the nCounter SPRINT system should have a range of 0.1 - 1.8 spots per square micron.

Positive Control Linearity QC

This metric performs a correlation analysis after Log(2) transformation of the expression values. The correlation is tested between the known concentrations of positive control target molecules added by NanoString and the resulting Log(2) counts. Correlation values lower than 0.95 may indicate an issue with the hybridization reaction and/or assay performance.

Limit of Detection QC

The limit of detection of the assay compares the positive control probes and the negative control probes. Specifically, it is expected that the 0.5 fM positive control probe (Pos_E) will produce raw counts that are at least two standard deviations higher than the mean of the negative control probes (represented by the box plot). The critical value for each sample is drawn as a red horizontal line for each sample.

Table of QC Flags

Table of Sample Annotations

PanCancer IO 360 Biological Signatures

Signature Introduction

Signatures are organized here in alphabetical order. They are color-coded by biology, similar to the color-coding in this image. Tumor signatures are listed in orange, Immune signatures in blue, and Micronenvironment signatures in green. Below each signature name is the signature category with which it is associated.

Biological Signatures

Signature Scales

Most scores can be interpreted on a log2 scale, with a unit increase in score corresponding to a doubling of its gene expression levels. The following scores occupy different scales:

  • APM loss, MMR loss, and JAK-STAT loss: these scores fall near zero in tumors without loss of expression events and increase sharply as any of their genes suffer expression loss.

    • Scores below 2 are consistent with typical expression levels (no loss).
    • Scores between 2 and 3 are borderline.
    • Scores above 3 imply a signature gene has much lower expression that would be expected.

TIS

Anti-Tumor Immune Activity

Tumor Inflammation Signature. TIS measures the abundance of a peripherally suppressed adaptive immune response within the tumor.

  • This signature is trained to predict response to anti-PD1 therapy (pembrolizumab). It consists of genes related to Interferon gamma signaling (IFNγ), antigen presentation, natural killer (NK) and T cells and inhibitory pathways. It also consists of normalization genes that have been selected to give consistent expression levels across most tissue or tumor types.
  • This signature is useful for predicting response to anti-PD1 therapy and determining hot and cold immune status across multiple cancer types.

APM

Tumor Immunogenicity

Antigen presenting (or processing) machinery. This signature measures the abundance of genes in the MHC Class I antigen presentation pathway and some key genes involved in processing the antigens prior to presentation. Typically, antigens from the cell cytoplasm are presented on Class I and recognized by the TCR on cytolytic CD8+ T cells. MHC Class I is expressed by all nucleated cells in the body, but downregulation of Class I MHC pathways is an evasion strategy that can be employed by tumor cells. An effective anti-tumor immune response depends on cytolytic T cells encountering neoantigens presented on the tumor cell surface. Strong anti-tumor immune responses are typically accompanied by high expression of antigen presentation genes.

APM Loss

Tumor Immunogenicity

Antigen presenting (or processing) machinery expression loss. This signature measures the extent to which any of several key major histocompatibility complex (MHC) genes have atypically low expression conditional on total MHC Class I gene expression. Values below 2 are expected from tumors with intact antigen presenting machinery. Antigen presentation via MHC class I in tumor cells is a major mechanism for immune recognition of tumors. Mutation or loss of expression of key class I MHC genes has been observed to confer resistance to immunotherapy.

Apoptosis

Tumor Sensitivity to Immune Attack

This signature captures genes associated with apoptotic processes, specifically with genes involved in mitochondrial membrane integrity. It includes both pro- and anti-apoptotic genes.

ARG1

Inhibitory Immune Signaling

Arginase-1 gene expression. ARG1 is expressed by myeloid cells and catalyzes the conversion of arginine to ornithine and urea. This suppresses T cell responses by preventing proliferation, which is dependent upon the availability of arginine for protein synthesis.

B Cells

Immune Cell Population Abundance

This signature measures the abundance of B cells in the tumor microenvironment. B cells are one of the main components of the humoral immune response. B cells have many functions including developing into plasma cells that secrete antibodies and presenting antigens to T cells. Presence of B cells in the tumor microenvironment have been shown to be associated with tertiary lymphoid structures in some tumor types.

B7-H3

Inhibitory Tumor Mechanisms

B7-H3 (CD276) gene expression. B7-H3 is a negative regulator of T cell activity that is expressed on both tumor and immune cells.

CD45

Immune Cell Population Abundance

CD45 gene expression. CD45 is a protein tyrosine phosphatase, receptor type C (PTPRC) and is expressed on all immune cells.

CD8 T Cells

Immune Cell Population Abundance

This signature measures the abundance of CD8+ T cells in the tumor microenvironment.

CTLA4

Inhibitory Immune Signaling

Cytotoxic T-lymphocyte-associated protein 4 gene expression. CTLA4 is a checkpoint molecule that inhibits T cell priming by competitively binding CD80/86 to prevent co-stimulation of via CD28.

Cytotoxic Cells

Immune Cell Population Abundance

This signature measures the abundance of cytotoxic cells in the tumor microenvironment. Cytotoxic cells such as natural killer (NK) and CD8+ T cells use a number of molecules, including perforin, granzymes and killer cell lectin-like receptor (KLRG) family members to recognize, penetrate and kill infected cells. Cytotoxic activity is the mechanism by which the immune system most effectively kills tumor cells.

Cytotoxicity

Anti-Tumor Immune Activity

This signature measures the molecules used by natural killer (NK) and CD8+ T cells to mount a cytolytic attack on tumor cells. Cytotoxic cells such as NK and CD8+ T cells, use a number of molecules, including perforin, granzymes and granulysin to penetrate and kill infection cells and tumors. Cytotoxic activity is the mechanism by which the immune system most effectively kills tumor cells.

DC

Immune Cell Population Abundance

Dendritic cell abundance. This signature measures the abundance of dendritic cells in the tumor microenvironment.

Endothelial Cells

Stromal Factors

This signature measures genes associated with vascular tissue and angiogenesis. Angiogenesis is important for nutrient trafficking to the tumor and proper oxygenation for tumor growth. Tumor angiogenesis forms leaky inefficient vessels that can reduce efficiency of lymphocyte trafficking to tumors.

Exhausted CD8

Immune Cell Population Abundance

This signature measures the abundance of exhausted CD8 cells in the tumor microenvironment. T cells in the tumor microenvironment often become less functional (exhausted) after exposure to factors from suppressive immune, stromal, and malignant cells. These exhausted T cells lose their ability to eliminate tumors. Exhausted T cells express markers and genes that are associated with reduced anti-tumor activity. This signature looks at the functional activity of T cells in the tumor microenvironment.

Glycolytic Activity

Inhibitory Metabolism

This signature measures genes participating in energy consumption. Up-regulated glycolysis and corresponding increased glucose consumption is nearly universal in tumors. Glycolysis may inhibit effective immune responses by depriving immune cells of glucose in the tumor microenvironment and changing other molecules that would ordinarily signal the immune system.

Hypermutation

Tumor Immunogenicity

This signature measures downstream gene expression changes associated with loss of MMR activity in tumor types where MMR loss is more prevalent.

Hypoxia

Inhibitory Metabolism

This signature measures genes associated with reduced oxygenation in the tumor. Hypoxia can induce expression of many cancer promoting processes (e.g. invasion, motility, metabolic reprogramming) and can promote resistance to immune cell-mediated cytolysis and reduced cytolytic activity in natural killer (NK) and CD8+ T cells.

IDO1

Inhibitory Tumor Mechanisms

Indoleamine 2,3-dioxygenase 1 gene expression. IDO1 is expressed by tumor, immune, and stromal cells and is the rate-limiting enzyme of tryptophan catabolism. By catalyzing the degradation of tryptophan, which is necessary for cytolytic T cell proliferation and activity, IDO1 inhibits anti-tumor immune responses.

IFN Downstream

Anti-Tumor Immune Activity

Interferon Signaling Response. This gene signature reflects activation of a broad set of interferon signaling pathways. Both type I and II interferons are implicated in tumor immune responses and regulate anti-tumor activity. Malignant and immune cells in the tumor microenvironment produce type 1 interferents, which have been shown to play a role in immunosurveillance. Interferon signaling is associated with improved patient outcome.

IFN Gamma

Anti-Tumor Immune Activity

Interferon gamma signaling. This signature tracks the canonical response to type II interferon, including the most universal components of that response. IFNγ induces macrophage and natural killer (NK) cell activation, increases antigen presentation, and induces gene transcription patterns that can lead to immune cell recruitment to the tumor. IFNγ signaling expression is associated with response to anti-PD1/L1 therapy.

IL10

Inhibitory Immune Signaling

Interleukin-10 gene expression. IL10 is a pleiotropic cytokine expressed predominately by monocytes. It impacts multiple aspects of the tumor immune response, including antigen presentation, T cell activation, and cytokine production.

Immunoproteasome

Tumor Immunogenicity

This signature measures key components of the immunoproteasome. The immunoproteasome is a specialized variant of the classical proteasome. It is assembled in immune cells and non-immune cells after exposure to proinflammatory cytokines or oxidative stress. The immunoproteasome induces different patterns of protein degradation, thus generating novel antigens that are presented in MHC class I. Immunoproteasome gene expression can be associated with increased tumor immunogenicity.

Inflammatory Chemokines

Inhibitory Immune Signaling

Inflammatory chemokines recruit both myeloid and lymphoid populations to the tumor microenvironment.

JAKSTAT Loss

Tumor Sensitivity to Immune Attack

JAK-STAT loss = JAK-STAT pathway expression loss. This signature measures loss of genes associated with JAK-STAT signaling, which has been identified as a mechanism of acquired resistance to immune checkpoint blockade.

Lymphoid

Anti-Tumor Immune Activity

Lymphoid compartment activity. This signature measures a broad set of genes involved in the functioning of lymphoid cells, including genes quantifying T cell abundance, B cell abundance, natural killer (NK) cell abundance, cytotoxic activity, interferon gamma signaling, JAK-STAT signaling, and T-cell co-stimulatory and co-inhibitory molecules. This signature captures a broad look at the lymphoid immune status of a tumor.

Macrophages

Immune Cell Population Abundance

This signature measures the abundance of macrophages in the tumor microenvironment. Macrophages can either augment tumor immunity (e.g. by presenting antigen) or suppress tumor immunity (e.g. by releasing immunosuppressive cytokines).

MAGEs

Tumor Immunogenicity

Melanoma-Associated Antigen Gene Expression. This signature measures several melanoma-associated antigens from the cancer testis (CT) antigen family. This family of CT antigens are expressed in a variety of cancers. They are important tumor-specific neoantigens that have been implicated in tumor biology and are often used as immunotherapy targets. This signature is useful to distinguish CT antigen expression before and after therapy.

Mast Cells

Immune Cell Population Abundance

This signature measures the abundance of mast cells in the tumor microenvironment.

MHC2

Anti-Tumor Immune Activity

Major histocompatibility complex class II antigen presentation. This signature measures the major human leukocyte antigens (HLA) involved in MHC Class II antigen presentation. Professional antigen presenting cells (dendritic cells, macrophages and B cells) use the class II MHC to present extracellular antigens to CD4+ T cells. Activation of CD4+ T cells induces expression of cytokines that can promote cytotoxic T cell activation and effective anti-tumor adaptive immune responses. Presence of MHC Class II molecules is associated with improved patient outcome.

MMR Loss

Tumor Immunogenicity

Mismatch Repair Gene Expression Loss. This signature measures the expression levels of several key mismatch repair genes. Mismatch repair deficiency often results when one of these genes has significant expression loss. Mismatch repair deficient tumors have high mutation rates due to the loss of important DNA-repair mechanisms. MMR loss is associated with better immune recognition in certain types of tumors. High scores in this signature indicate MMR deficiency. Mismatch repair deficiency and microsatellite instability (MSI) predict response to immune checkpoint blockade.

MSI Predictor

Tumor Immunogenicity

This is a signature that integrates the output of the MMR loss signature and the hypermutation signature to predict MMR activity loss.

Myeloid

Anti-Tumor Immune Activity

Myeloid compartment activity. This signature measures key marker and effector genes of myeloid lineage immune cells. Myeloid cells regulate both anti- and pro-tumor activities. This signature captures a broad look at the myeloid immune status of a tumor.

Myeloid Inflammation

Inhibitory Immune Signaling

Myeloid-derived inflammatory signaling. This signature measures gene expression for myeloid lineage cells with pro- and anti-tumor functions. Myeloid cells produce a number of cytokines and chemokines that promote a state of inflammation. Depending on the context, these can promote tumorigenesis or anti-tumor immune responses.

Neutrophils

Immune Cell Population Abundance

This signature measures the abundance of neutrophils in the tumor microenvironment. In the context of the tumor, neutrophils tend to have immunosuppressive activity.

NK CD56dim

Immune Cell Population Abundance

A subset of natural killer cells. This signature represents the subset of NK cells with the most cytolytic activity.

NK Cells

Immune Cell Population Abundance

Natural Killer cell abundance. This signature measures the abundance of NK cells in the tumor microenvironment.

NOS2

Inhibitory Immune Signaling

Nitric Oxide Synthase 2 gene expression. NOS2 is induced by Interferon gamma signaling (IFNγ). It regulates expression of nitric oxide, which at low levels can promote tumor growth but at high levels may be cytostatic or cytotoxic to tumor cells.

PD-1

Inhibitory Immune Signaling

Program cell death receptor 1 gene expression. Program cell death receptor 1 (PD-1, PDCD1, CD279) is expressed predominantly on lymphocytes. It is upregulated upon activation and becomes a negative regulator of activation by preventing proliferation and cytokine secretion. PD-1 expression has been shown to be associated with tumor-specific T cells.

PD-L1

Inhibitory Tumor Mechanisms

Program cell death ligand 1 gene expression. Program cell death ligand 1 (PD-L1, CD274) is a ligand for PD-1 and negative regulator of T cell activity that is expressed on both tumor and immune cells.

PD-L2

Inhibitory Immune Signaling

Program cell death ligand 2 gene expression. Program cell death ligand 2 (PD-L2, PDCDLG2, CD273) is a ligand for PD-1 and negative regulator of T cell activity that is expressed on antigen-presenting cells.

Proliferation

Tumor Sensitivity to Immune Attack

Tumor Proliferation. This signature measures genes involved in tumor proliferation. A highly proliferative tumor can overcome an immune response if replication exceeds immune mediated detection and elimination.

Stroma

Stromal Factors

Stromal Tissue Abundance. This signature measures stromal components in the tumor microenvironment. The tumor stroma is the collection of non-cancerous and nonimmune tissue components surrounding the tumor. Stroma can act as a physical barrier that excludes immune cells from the tumor, preventing effective anti-tumor immunity even when tumor-associated antigens have induced immune cell priming and activation. These cells can also secrete important signals to the tumor, affecting tumor biology and response to the immune system.

T Cells

Immune Cell Population Abundance

This signature measures the abundance of T cells in the tumor microenvironment.

TGF-Beta

Inhibitory Tumor Mechanisms

Transforming Growth Factor Beta gene expression. TGFβ (TGFB1) is a pleotropic cytokine which inhibits anti-tumor immune activity and promotes tumor growth and survival.

TH1 Cells

Immune Cell Population Abundance

T-box transcription factor TBX21 (T-bet) expressing cell abundance. T-bet is the canonical transcription factor that defines TH1 T cells and is used to measure TH1 cell abundance. TH1 T cells promote anti-tumor immune activity (particularly supporting CD8+ T cell function) by producing Interferon gamma (IFNγ).

TIGIT

Inhibitory Immune Signaling

T cell immunoreceptor and Ig and ITIMS gene expression. T cell immunoreceptor and Ig and ITIMS domains (TIGIT) is an immune checkpoint molecule that suppresses anti-tumor immune activity in CD8+ T cells and NK cells.

Treg

Immune Cell Population Abundance

Regulatory T cell abundance. Treg is measured by gene expression of Forkhead box P3 (FOXP3). FOXP3 is the canonical transcription factor that defines the regulatory T cell (Treg) population and is used to measure Treg abundance. Regulatory T cells suppress other T cell activities through a variety of mechanisms.

Selected Publications

Ayers, Mark, et al. IFN-γ–related mRNA profile predicts clinical response to PD-1 blockade. The Journal of Clinical Investigation 127.8 (2017).

Danaher, Patrick, et al. Gene expression markers of Tumor Infiltrating Leukocytes. Journal for immunotherapy of cancer 5.1 (2017): 18.

Danaher P, Warren S, Ong S, Elliott N, Cesano A, Ferree S. A gene expression assay for simultaneous measurement of microsatellite instability and anti-tumor immune activity. Journal for immunotherapy of cancer 7.1 (2019):15.

Satoh, Jun-ichi, and Hiroko Tabunoki. A comprehensive profile of ChIP-Seq-based STAT1 target genes suggests the complexity of STAT1-mediated gene regulatory mechanisms. Gene regulation and systems biology 7 (2013): 41.

Becht, Etienne, et al. Estimating the population abundance of tissue-infiltrating immune and stromal cell populations using gene expression. Genome biology 17.1 (2016): 218.

Spranger, Stefani, Riyue Bao, and Thomas F. Gajewski. Melanoma-intrinsic β-catenin signalling prevents anti-tumour immunity. Nature 523.7559 (2015): 231.

Harris, B. H. L., et al. Gene expression signatures as biomarkers of tumour hypoxia. Clinical Oncology 27.10 (2015): 547-560.

Manson, G., et al. Biomarkers associated with checkpoint inhibitors. Annals of Oncology 27.7 (2016): 1199-1206.

Blank, Christian U., et al. The “cancer immunogram”. Science 352.6286 (2016): 658-660.

Methods

Normalization

Normalization in the 360 reports differs from normalization in nSolver. The goal is to adjust for cartridge differences using a panel standard containing all 360 gene targets that is run on each cartridge within the experiment such that comparisons can be made between the scores across batches. Normalization takes place in two steps. The first step differs depending on whether the genes are in the TIS signature, or not, and is described below. Zero counts on the raw scale are converted to ones prior to normalization.

Housekeeper Normalization for non-TIS Genes

Genes are normalized using a ratio of the expression value to the geometric mean of all housekeeping genes on the panel.

Housekeeper Normalization for TIS Genes

Genes in the TIS signature are normalized using a ratio of the expression value to the geometric mean of the housekeeper genes used only for the TIS signature.

Panel Standard Normalization

Genes are additionally normalized using a ratio of the housekeeper-normalized data and a panel standard run on the same cartridge as the observed data. In the absence of a panel standard column, values from panel standard run on the same codeset lot as the observed data may be substituted.

Final Adjustments

The housekeeper-normalized and panel standard-normalized data is Log(2) transformed. A constant of 8 is added to TIS so that it is on the same scale as investigational use only (IUO) TIS, making scores comparable across research use only (RUO) and IUO assays. Other non-TIS signatures are also adjusted with constants to express values in a similar range.

Differential Expression Analysis

Grouping Variable

Differential expression is fit on a per gene or per signature basis using a linear model for analyses without a blocking factor. The statistical model uses the expression value or signature score as the dependent variable and fits a grouping variable as a fixed effect to test for differences in the levels of that grouping variable.

Expression(gene or signature)= μ+Group+ε

P-values are adjusted within each analysis, gene or signature, and on the grouping variable level difference t-test using the Benjamini and Yekutieli False Discovery Rate (FDR) adjustment to account for correlations amongst the tests. All models are fit using the limma package in R.

Survival Analysis

Grouping Variable

If a grouping variable is present, the survival analysis used to create the forest plot incorporates a proportional hazards model with the survival outcome as a dependent variable, the observed normalized gene expression or signature score data as a continuous covariate, and the grouping variable included as a strata variable in the model which results in the model being a frailty model.

Survival(time,event)= μ+Expression_(gene or signature)+Group+ ε

The analysis method is performed on a by gene or by signature basis, as appropriate, and uses the regression routines implemented in the R package survival. All p-values are adjusted for the number of tests within each type of analysis (gene or signature) using the Benjamini and Yekutieli False Discovery Rate (FDR) method to account for correlations amongst the tests.

There are no Kaplan-Meier curves available for frailty models and thus are not present when the analysis is fit as a frailty model with a random effect.

Time Series Analysis

Differential expression is fit on a per gene or per signature basis using a mixed linear model for analyses with subject as a blocking factor to account for the temporal effects in the model. The statistical model uses the expression value or signature score as the dependent variable and fits a grouping variable as a fixed effect to test for differences in the levels of that grouping variable. The duplicateCorrelation function within the limma R package is used to assess the correlation between subsequent time points. This correlation estimate is fit into the linear mixed effect model with subject as the random effect and the correlation between the repeated temporal measurements.

Expression_(gene or signature)= μ+SubjectID+Group+ε

P-values are adjusted within each analysis, gene or signature, and on the grouping variable level difference t-test using the Benjamini and Yekutieli False Discovery Rate (FDR) adjustment to account for correlations amongst the tests. All models are fit using the limma package in R.

References

Benjamini Y and Yekutieli D. 2001. The control of the false discovery rate in multiple testing under dependency. Annals of Statistics 29:4:1165-1188 https://projecteuclid.org/euclid.aos/1013699998

Ritchie ME, Phipson B, Wu D, Hu Y, Law CW, Shi W, Smyth GK (2015). limma powers differential expression analyses for RNA-sequencing and microarray studies. Nucleic Acids Research, 43(7), e47. https://bioconductor.org/packages/release/bioc/html/limma.html

Therneau TM (2015). A Package for Survival Analysis in S. version 2.38, https://CRAN.R-project.org/package=survival.

Therneau TM and Grambsch PM (2000). Modeling Survival Data: Extending the Cox Model. Springer, New York. ISBN 0-387-98784-3.