Pan-Cancer Mapping of the Tumor Immune Landscape through Metagene Clustering and Predictive Modeling

Soham Chatterjee

q-bio.GNMar 28, 2026

Pan-Cancer Mapping of the Tumor Immune Landscape through Metagene Clustering and Predictive Modeling

Authors: Soham Chatterjee

Abstract

As immunotherapies become standard cancer treatments, it is increasingly important to identify a patient's immune profile, which encompasses the activity of immune cells within the tumor microenvironment and the presence of specific biomarkers. However, we lack mechanistic explanations drivers of immune phenotypes. Despite advances in immune profiling with high-throughput sequencing, the mechanisms driving them remain unclear. This study aimed to identify novel, robust immune-related gene clusters (metagenes) and evaluate their prognostic significance and functional relevance across various pan-cancer types using a comprehensive computational pipeline. We acquired pan-cancer bulk RNA-seq and established immune subtypes from The Cancer Genome Atlas (TCGA). Using expression-based filtering and clustering of genes with ANOVA and Gaussian Mixture Model (GMM), we identified 48 unique metagenes. These metagenes achieved 87% accuracy in predicting the established subtypes. SHAP analysis revealed the most predictive metagenes per subtype, while functional enrichment analysis identified their associated pathways. Genes were ranked by differential expression between high- and low-expression groups. The metagenes revealed insights, including co-expression of immune activation and regulatory factors, links between cell cycle regulation and immune evasion, and dynamic microenvironment remodeling signatures. Kaplan-Meier survival analysis and multivariate Cox Regression revealed that many metagenes had prognostic value for overall survival. Overall, the metagenes represent coordinated biological programs across diverse cancer types, providing a foundation for developing robust, broadly applicable immuno-oncology biomarkers that extend beyond single-gene markers. They demonstrate prognostic value across cancer types and hold potential to guide immunotherapy treatment decisions.

Figures & tables

Subtype	Number of Patient Samples
C1	2416
C2	2591
C3	2397
C4	1157
C5	385
C6	180

Table 1: Distribution of patient samples across subtypes

Model	Features	Metagenes	Hyperparameter	Value
SVM	46	36	C	2.2165
			Degree	4
Random Forest	56	38	n_estimators	188
			Max depth	10
XGBoost	68	48	n_estimators	331
			Max depth	5

Table 2: Summary of machine learning models, including RFECV-selected features, number of unique metagenes among selected features, and hyperparameters optimized through tuning.

Panel of Figure 1 — Figure 1: Gene selection and clustering analysis. (a) t-SNE visualization of ANOVA-selected genes, and (b) t-SNE visualization using XGBoost-selected features.

Metric	XGB Train	XGB Test	SVM Train	SVM Test	RF Train	RF Test
Accuracy	100.0% $\pm$ 0.0%	87.8% $\pm$ 0.9%	97.8% $\pm$ 0.1%	86.5% $\pm$ 0.5%	92.1% $\pm$ 0.2%	82.5% $\pm$ 1.1%
Precision (Macro)	100.0% $\pm$ 0.0%	82.9% $\pm$ 2.1%	98.1% $\pm$ 0.2%	78.6% $\pm$ 1.4%	87.3% $\pm$ 0.5%	75.7% $\pm$ 1.7%
Precision (Weighted)	100.0% $\pm$ 0.0%	87.4% $\pm$ 0.9%	97.8% $\pm$ 0.1%	86.3% $\pm$ 0.5%	93.0% $\pm$ 0.1%	83.4% $\pm$ 1.2%
Recall (Macro)	100.0% $\pm$ 0.0%	78.1% $\pm$ 1.2%	97.1% $\pm$ 0.2%	77.8% $\pm$ 1.1%	94.1% $\pm$ 0.2%	77.3% $\pm$ 2.2%
Recall (Weighted)	100.0% $\pm$ 0.0%	87.8% $\pm$ 0.9%	97.8% $\pm$ 0.1%	86.5% $\pm$ 0.5%	92.1% $\pm$ 0.2%	82.5% $\pm$ 1.1%
F1 Score (Macro)	100.0% $\pm$ 0.0%	79.7% $\pm$ 1.5%	97.6% $\pm$ 0.2%	78.1% $\pm$ 1.2%	89.4% $\pm$ 0.5%	75.9% $\pm$ 1.8%

Table 3: Comparison of training and test set metrics across the three models.

Figure 2: Normalized confusion matrix of predicted versus true subtype labels for the XGBoost model.

Rank	Metagene	# Subtypes / # Appearances
1	Metagene 82	4 / 8
2	Metagene 147	3 / 3
3	Metagene 87	3 / 3
4	Metagene 27	2 / 3
5	Metagene 59	2 / 2
6	Metagene 55	2 / 2

Table 4: Top ten most predictive metagenes from SHAP analysis. The number of subtypes in which a metagene appeared and the total number of appearances across summary statistics were calculated, considering only the top five metagene-derived features per subtype.

Panel of Figure 3 — Figure 3: Panels a–f (corresponding to subtypes C1–C6) show SHAP value distributions for the top 20 most predictive metagene-derived features identified for each subtype. Each violin represents the contribution of a feature to the model’s prediction, with wider sections indicating higher sample density. Positive and negative SHAP values reflect feature influence toward or against classification of the corresponding subtype.

Metagene	Stat	Low (%)	High (%)	$\Delta$ 5-yr (%)	Adj. p
Metagene 110	mean	50.9	69.3	18.4	2.38e-64
Metagene 110	median	51.0	69.2	18.2	2.37e-57
Metagene 81	mean	51.0	68.9	17.9	2.68e-53
Metagene 101	mean	50.9	68.8	17.9	1.98e-52
Metagene 14	mean	51.8	68.5	16.7	6.33e-50
Metagene 107	median	52.2	68.0	15.8	2.29e-47

Table 5: Top metagene-derived features ranked by survival significance. The table reports 5-year survival percentages for each group, the absolute survival difference, and the adjusted p-value from Kaplan–Meier analysis (ranked by lowest adjusted p-value).

Panel of Figure 4 — Figure 4: Kaplan–Meier plots showing overall survival differences between high (above-median) and low (below-median) expression groups for three distinct metagene-derived features: (a) Metagene 110 (mean), (b) Metagene 81 (mean), and (c) Metagene 101 (mean).

Metagene	Stat	HR	95% CI	Adj. p
Metagene 75	mean	1.099	1.042 - 1.160	3.84e-03
Metagene 14	mean	1.051	1.021 - 1.083	5.89e-03
Metagene 107	median	1.055	1.022 - 1.089	5.89e-03
Metagene 107	mean	1.055	1.021 - 1.090	7.29e-03
Metagene 78	mean	1.081	1.023 - 1.142	2.76e-02
Metagene 133	median	0.949	0.915 - 0.986	2.76e-02

Table 6: Hazard ratios for top metagene-derived features. The table reports hazard ratios (HR) with 95% confidence intervals and adjusted p-values from Cox regression.

Rank	Metagene	Biological Pathways (Min adj p-value)
1	Metagene 21	Cell cycle process (6.4e-37)
2	Metagene 82	Defense response to virus (1.1e-35)
3	Metagene 114	T-cell activation / MHC I binding (9.2e-32)
4	Metagene 107	Cell cycle process (3.4e-30)
5	Metagene 85	Cell cycle process (4.0e-26)
6	Metagene 15	T-cell activation (3.9e-24)

Table 7: Top metagenes ranked by minimum adjusted p-value and their associated biological pathways.

Metagene	Num. Genes	Top Genes	Expr. (Mean)
Metagene 147	51	HLA-DQA1	4.52
		IL2RA	4.43
		HLA-DQA2	4.38
Metagene 82	41	OASL	4.27
		ZBP1	4.16
		IFI44L	4.03

Table 8: Representative metagenes with top upregulated genes and mean expression values.

Cancer Subtype Classification

See all 14 papers

Mar 17, 2026

HistoAtlas: A Pan-Cancer Morphology Atlas Linking Histomics to Molecular Programs and Clinical Outcomes

We present HistoAtlas, a pan-cancer computational atlas that extracts 38 interpretable histomic features from 6,745 diagnostic H&E slides across 21 TCGA cancer types and systematically links every feature to survival, gene expression, somatic mutations, and immune subtypes. All associations are covariate-adjusted, multiple-testing corrected, and classified into evidence-strength tiers. The atlas recovers known biology, from immune infiltration and prognosis to proliferation and kinase signaling, while uncovering compartment-specific immune signals and morphological subtypes with divergent outcomes. Every result is spatially traceable to tissue compartments and individual cells, statistically calibrated, and openly queryable. HistoAtlas enables systematic, large-scale biomarker discovery from routine H&E without specialized staining or sequencing. Data and an interactive web atlas are freely available at https://histoatlas.com .

Pierre-Antoine Bannier

Computational Pathology Whole-Slide Image Analysis Cancer Survival Prediction Scientific Discovery

Ask PDF

Apr 17, 2026

Graph Transformer-Based Pathway Embedding for Cancer Prognosis

Accurate prediction of cancer progression remains a challenge due to the high heterogeneity of molecular omics data across patients. While biologically informed models have improved the interpretability of these predictions, a persistent limitation lies in how they encode individual genes to construct pathway representations. Existing hierarchical models typically derive gene features by directly mapping raw molecular inputs, whereas integration frameworks often rely on simple statistical aggregations of patient-level signals. These approaches often fail to explicitly learn a shared base representation for each gene, thereby limiting the expressiveness and biological accuracy of downstream pathway embeddings. To address this, we introduce PATH, a modulation-based, patient-conditioned gene embedding strategy. PATH represents a paradigm shift by starting from a shared base embedding for each gene, preserving a stable biological identity across the population, and then dynamically adapting it using patient-specific copy number variation (CNV) and mutation signals. This allows the model to capture subtle individual molecular variations while maintaining a consistent latent understanding of the gene itself. We integrate PATH into a graph transformer framework that models interactions among biologically connected pathways through pathway-guided attention. Across pancancer metastasis prediction, PATH achieves an F1 score of 0.8766, representing an 8.8 percent improvement over the current SOTA multi-omics benchmarks. Beyond superior predictive accuracy, our approach identifies biologically meaningful pathways and, crucially, reveals disease-state-specific pathway rewiring, offering new insights into the evolving pathway-pathway interactions that drive cancer progression.

Koushik Howlader, Md Tauhidul Islam, Wei Le

Iowa State University, Iowa, USA · Stanford University, CA, USA

Cancer Genomics Cancer Survival Prediction Clinical Outcome Prediction Oncology

Ask PDF

Apr 27, 2026

PathMoG: A Pathway-Centric Modular Graph Neural Network for Multi-Omics Survival Prediction

Cancer survival prediction from multi-omics data remains challenging because prognostic signals are high-dimensional, heterogeneous, and distributed across interacting genes and pathways. We propose PathMoG, a pathway-centric modular graph neural network for multi-omics survival prediction. PathMoG reorganizes genome-scale inputs into 354 KEGG-informed pathway modules, introduces a Hierarchical Omics Modulation module to condition gene-expression representations on mutation, copy number variation, pathway, and clinical context, and uses dual-level attention to capture both intra-pathway driver signals and inter-pathway clinical relevance. We evaluated PathMoG on 5,650 patients across 10 TCGA cancer types and observed consistent improvements over representative survival baselines. The framework further provides gene-level, pathway-level, and patient-level interpretability, supporting biologically grounded and clinically relevant risk stratification.

Di Wang, Chupei Tang, Junxiao Kong +3

1Cuiying Honors College, Lanzhou University, China · School of Mathematics and Statistics, Lanzhou University, China · School of Mathematics and Statistics, Lanzhou University, China and Shanghai Innovation Institute +1

Survival Analysis Graph Neural Networks Cancer Survival Prediction Cancer Genomics Oncology Modular Neural Networks+1

Ask PDF

Abstract

Cancer Subtype Classification

Explore similar work

HistoAtlas: A Pan-Cancer Morphology Atlas Linking Histomics to Molecular Programs and Clinical Outcomes

Graph Transformer-Based Pathway Embedding for Cancer Prognosis

PathMoG: A Pathway-Centric Modular Graph Neural Network for Multi-Omics Survival Prediction