https://orcid.org/0009-0009-8093-3436
Figures
Abstract
Intratumor heterogeneity (ITH) is a critical factor influencing cancer progression, therapeutic response, and the development of drug resistance. Despite its importance, the lack of a definitive gold standard for ITH quantification has hindered consistent clinical application. To bridge this gap, we developed ITHindex, a web-based, research-enabling platform that integrates a pragmatic subset of 17 user-accessible ITH algorithms within the R Shiny framework. The tool supports diverse data modalities, including somatic mutation, copy number variation, transcriptomic, proteomic, and methylation profiles. By analyzing 11,242 samples across 32 cancer types from The Cancer Genome Atlas (TCGA) and 4,904 samples from cBioPortal, alongside 398 paired tissue and plasma samples, we validated the platform’s utility in quantifying ITH and characterizing the relationships between diverse metrics. ITHindex streamlines the computational pipeline, providing researchers with a robust tool for systematic ITH investigation and data-driven biomarker discovery. The ITHindex server is freely accessible at https://shinyapps.brbiotech.com/app/ithindex.
Author summary
Prior knowledge indicated that ITH mirrors the complexity of tumor clones and holds promise as a predictive biomarker for the efficacy of immunotherapy. Several algorithms have been developed to quantify ITH based on omics data. However, implementing these algorithms often involves multiple software tools, advanced programming skills, and complex workflows, which can hinder accessibility and reproducibility. We introduced ITHindex, an integrated platform that integrated these algorithms to streamline the practice of ITH quantification for clinical researchers. ITHindex combines these algorithms into a user-friendly platform, seamlessly integrating computational methods with proprietary modules through an intuitive graphical interface. By integrating modality-specific algorithms, multi-omics data, and customizable parameters, the platform empowers researchers to conduct ITH quantification without coding. This platform reduces technical barriers, enhances reproducibility, and fosters collaborative, data-driven biomarker discovery.
Citation: Liu Y, Jiang Y, Xie W, Duan F, Fu S, Zhao J, et al. (2026) ITHindex: An integrated web-based platform for intratumor heterogeneity evaluation. PLOS Digit Health 5(8): e0001654. https://doi.org/10.1371/journal.pdig.0001654
Editor: Donnella S. Comeau, Beth Israel Deaconess Medical Center, UNITED STATES OF AMERICA
Received: September 22, 2025; Accepted: July 23, 2026; Published: August 28, 2026
Copyright: © 2026 Liu et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Data Availability: The public datasets analyzed in this study are available from TCGA, GEO, and cBioPortal. All code and example data for ITHindex are publicly available at https://github.com/yuchenstar/ITHindex. The in-house dataset is not publicly available due to institutional and national data-protection policies regarding potentially identifiable genomic information. Anonymized data may be made available upon reasonable request, subject to ethical and administrative approvals. Inquiries and data-access requests should be directed to info@brbiotech.com with a clear statement of the research purpose.
Funding: This work was supported by the Capital’s Funds for Health Improvement and Research (2024-2-4029 to YT. Liu) and the National Natural Science Foundation of China (82472888 to YT. Liu). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Competing interests: I have read the journal’s policy and the authors of this manuscript have the following competing interests: Wenchuan Xie, Jinyu Yang, Yuzi Zhang, Guoqiang Wang, and Feidie Duan are current employees of Burning Rock Biotech. Jing Zhao and Shangli Cai were employed by Burning Rock Biotech during the conduct of this study.
Introduction
While originating from a single rogue somatic cell, a tumor commonly evolves into a complex ecosystem consisting of billions of cells exhibiting significant genetic and non-genetic variations, a phenomenon referred to as intratumor heterogeneity (ITH) [1]. Beyond traditional genomic aberrations, ITH is increasingly recognized as a multidimensional phenomenon that profoundly influences the tumor microenvironment (TME) and interacts with the intratumoral microbiome [2,3]. These non-genomic layers of heterogeneity — manifest in cellular morphology, metabolism, immune infiltration, and microbial diversity — create distinct ecological niches that drive disease progression, immune evasion, and treatment resistance [4–6]. Consequently, understanding this multidimensional landscape is essential for developing effective cancer therapies By recognizing and integrating these diverse facets of ITH, researchers and clinicians can work towards personalized strategies that target specific characteristics of cancer subgroups, ultimately improving patient outcomes.
Numerous metrics have been developed to assess ITH using genetic, transcriptomic, proteomic, and epigenetic data. Major metrics for genetic ITH (gITH) include Jaccard Similarity index [7], Mutant-Allele Tumor Heterogeneity (MATH) score [8], EXPANDS [9], PhyloWGS [10], EstimateClonality [11], Shannon entropy [12], and Copy number heterogeneity (CNH/CNHplus) [13]. For transcriptomic ITH (tITH), specific metrics include network-based Jensen-Shannon Divergence (nJSD) [14], standard deviations (SD) of absolute z-scored transcriptome levels (DEPTH2) [15] and Shannon entropy [16]. Proteomic ITH (pITH) and epigenetic data (epiITH) have fewer dedicated metrics. Among these, MATH, as one of the earliest algorithms, has been demonstrated to predict poor prognosis in head and neck squamous cell carcinomas, uterine corpus endometrial carcinoma, lung adenocarcinoma, breast cancer, and colorectal cancer based on elevated MATH-derived ITH scores [17]. Similarly, other algorithms have also been validated for their role in predicting prognosis across different cancer types [17]. In addition to prognosis prediction, Shannon entropy-based blood ITH has revealed that patients with lower ITH metrics experience notable overall survival (OS) and progression-free survival (PFS) benefit from immune checkpoint blockade therapies compared with chemotherapy [18].
Despite the abundance of available metrics and the promise for guiding treatment in clinical settings, a significant challenge remains that there is often a marked discordance among different ITH algorithms [19]. Because these methodologies rely on distinct statistical foundations—ranging from allele frequency distribution to phylogenetic tree reconstruction—results can vary substantially even when analyzing the same patient cohort [17]. Such inconsistencies hinder the establishment of standardized clinical thresholds and limit the reproducibility of ITH-based biomarkers across different studies. Furthermore, the current metrics are formulated using a range of tools, including those in R language, MATLAB, and some lacking calculation scripts entirely. The lack of a unified analytical framework makes it difficult for researchers to perform systematic comparisons or cross-validation.
To address this gap, we developed ITHindex, an interactive platform designed to enhance methodological rigor in tumor heterogeneity research. Researchers can upload their omics data to ITHindex and select from a range of metrics to compute relevant ITH scores. Beyond mere computational convenience, ITHindex facilitates the acquisition of multi-algorithmic perspectives, enabling researchers to evaluate the comparability of various metrics and ensuring that biological conclusions are robust rather than algorithm-dependent. By integrating diverse omics layers into a standardized pipeline, ITHindex aims to bridge the gap between complex evolutionary theory and reproducible precision oncology.
Results
Overview of ITHindex
ITHindex is an open-access web platform developed to integrate a pragmatic subset of user-accessible algorithms for assessing ITH. The current version of ITHindex (v0.2.4) integrates a curated selection of 17 algorithms derived from 14 seminal publications, which were filtered through specific inclusion and exclusion criteria to provide a practical toolset of Level 3 data-compatible methods across genomic (n = 8), transcriptomic (n = 5), proteomic (n = 2), and epigenetic (n = 2) modalities (S1 Table, S2A Fig). In terms of the number of sampling regions, 3 metrics involved multi-region sampling, while the remaining articles focused on single-region sampling. Regarding sample processing for detection, one metric utilized single-cell sequencing, while the others employed bulk DNA or RNA detection methods. From the perspective of algorithmic principles, all metrics were categorized into two types: statistical (descriptive and spatial) and information theory. Statistical algorithms are based on various statistical measures or methods to evaluate ITH, including similarity measure, mean absolute deviation (MAD), ratio, standard deviation, distance measures, and principal component analysis (PCA). The algorithm, grounded in information theory principles, utilized Shannon entropy to calculate ITH. The final output results are contingent on users’ selections of omics data and calculation methodologies.
ITHindex has already generated results for a total of 11,242 samples from TCGA utilizing 10 distinct metrics, which are downloadable as CSV files from the platform. Users can compute results for their own data, saving them locally in CSV format for subsequent research and analysis. Fig 1 provides an overview of the ITHindex tool’s framework.
Workflow and functional modules of ITHindex
ITHindex is powered by Shiny and offers a wide range of options for quantifying ITH. We have created a user-friendly web interface with the aim of incorporating multiple algorithms for ITH quantification, which are accessible through four main pages: “Home”, “ITH Index”, “Data” and “About”. The “Home” page provides an overview of ITHindex; the “ITH Index” page allows users to upload and analyze their own data; the “Data” section offers pre-analyzed ITH data from TCGA samples; and the “About” page includes four sections: “About ITHindex”, “Contact Us”, “Acknowledgments” and “FAQ (Frequently Asked Questions)”.
On the “ITH Index” tab, the ITH calculation process is segmented into four steps within an interactive collapsible box. A well-defined analysis workflow is illustrated in Fig 2. Initially, users can select the data type from four options provided by ITHindex: mutation data with variant frequency (VAF) or cancer cell fraction (CCF), copy number variation, mRNA expression/protein abundance, and methylation. Subsequently, users may choose an algorithm for quantifying ITH, with the option to customize specific parameters for certain algorithms. For example, when selecting “Mutation (VAF/CCF)” and “Shannon entropy”, users can adjust parameters such as Cancer Cell Fraction (Yes/No), Number of bins (10–100), Adjusted by max VAF/CCF (Yes/No), and Weighted to entropy (Yes/No) to suit their requirements. Users can tailor these parameters according to their preferences. Next, users can upload their data file from their local device through the browser tab. To facilitate optimal utilization of the tool, example data files for each data type in every algorithm are provided, and users are strongly encouraged to follow these examples when uploading their data. Upon uploading the data file, the subsequent tab, “View the upload data” displays the initial 10 rows and 20 columns of the uploaded data for verification. Finally, the tool presents the results table and plots after clicking “RUN!” button. ITHindex offers downloadable results and density plots for user access.
To ensure the robustness and reproducibility of ITH quantification, ITHindex implements a multi-layer validation pipeline. First, upon data upload, the platform performs real-time checks for input integrity, such as verifying column headers, detecting missing values, and validating data format compatibility with the chosen algorithm. Second, to facilitate reproducibility, ITHindex generates an “Analysis Parameters” box accompanying the results. This tab captures a comprehensive snapshot of the session, including the specific algorithm applied, all user-defined parameters, and the timestamp of the analysis. This documentation ensures that each computational step is transparent and verifiable, mitigating the risk of inadvertent parameter drift across analyses.
Comparison of various algorithms for ITH evaluation
To demonstrate the versatility and utility of ITHindex, we performed three comparative analyses using both public and in-house datasets.
Case 1: Cross-algorithm consistency and metric correlation.
A critical challenge in ITH research is determining whether outcome predictions are dependent on the choice of algorithm. ITHindex streamlines this by enabling rapid, multi-algorithm quantification.
First, we analyzed multi-regional bulk mRNA expression data from a lung cancer cohort (GSE33532). Using the two ITH metrics, scaled mean pairwise distance (sMPD) and diversity score, we identified a strong positive correlation (r = 0.99, P < 0.001; Fig 3A). Consistent results were observed across breast, liver, and cervical cancer datasets (Fig 3B, S2B-C Fig). The observed high degree of correlation is likely attributable to their shared theoretical foundations, as both algorithms are distance-based metrics designed to quantify divergence in expression profiles. Second, we compared Shannon entropy with MATH for mutation data across different sequencing strategies. Regardless of whether the data originated from targeted NGS panels or Whole-Exome Sequencing (WES), these two metrics demonstrated significant positive correlations in 13 of 16 cancer types (all P < 0.05, r > 0.3; Fig 3C-3E). Notably, however, for plasma-derived samples, significant correlations were only maintained in a subset of tumor types, specifically BRCA, RCC, and BTC (Fig 3D). Furthermore, we compared Shannon entropy with CNHplus using TCGA copy number variation (CNV) profiles. Our results showed that these two metrics were significantly correlated in most malignancies, encompassing 29 of 32 cancer types (all P values < 0.05, r > 0.3, Fig 3F). Last, when analyzing mRNA expression and methylation data, the correlation between Shannon entropy and SD-based ITH metrics was generally low, reaching significance only in HNSC and LIHC (Fig 3G-3H). These findings underscore that while algorithms with similar mathematical rationales yield highly concordant results, metrics derived from different omics layers or distinct statistical frameworks often capture divergent biological facets of ITH. This highlights the necessity of the multi-algorithmic approach provided by ITHindex.
(A) The correlation between sMPD and diversity score for the mRNA expression data.
in lung cancers; (B) The correlation between sMPD and diversity score for the mRNA expression data in breast cancers; (C) The correlation between Shannon entropy and MATH for mutation data derived from NGS panel data of tissue samples; (D) The correlation between Shannon entropy and MATH for mutation data derived from NGS panel data of plasma samples; (E) The correlation between Shannon entropy and MATH for tissue WES mutations data; (F) The correlation between Shannon entropy and CNHplus for tissue copy number data; (G) The correlation between Shannon entropy and standard deviations for tissue RNAseq data; (H) The correlation between Shannon entropy and standard deviations for tissue methylation data. (I) The ITH index correlations of Shannon entropy between WES and NGS panel; (J) The ITH index correlations of MATH between WES and NGS panel; (K) The ITH index correlations of data from tissue with plasma sample for Shannon entropy; (L) The ITH index correlations of data from tissue with plasma sample for MATH.
Case 2: Benchmarking NGS panels against WES.
To evaluate the reliability of the integrated algorithms, we conducted a systematic performance assessment focusing on platform heterogeneity, sequencing depth, and sample size. MATH and Shannon entropy were chosen as representative metrics for this validation due to their widespread application in clinical research [18,20–22]
First, we utilized 46 paired WES and NGS panel samples to evaluate the performance of Shannon entropy and MATH. The results showed the MATH-based ITH between the NGS panel and WES was lower than Shannon entropy (r of Shannon entropy: 0.72, P < 0.001, r for MATH: 0.40, P = 0.006, respectively, Fig 3I-3J). Second, genomic data from two LUAD cohorts were used to assess the stability of MATH and Shannon entropy across varied sequencing platforms and depths. As shown in Table 1 and S3 Fig, the ctDNA panel detected a wider range of VAFs (MAD: 0.009 for the ctDNA panel vs. 0.120 for WES, Table 1), and MATH-based ITH scores derived from the ctDNA panel were significantly higher than those from WES (median: 0.000 vs. 44.232, P < 0.05, S3 Fig). A consistent trend was observed in the comparison between the ctDNA panel and IMPACT (S3 Fig). Furthermore, although the MSK-IMPACT assay yielded a VAF MAD comparable to, or even higher than, that of WES (MAD: 0.156 for IMPACT vs. 0.132 for WES), the MATH score derived from IMPACT was significantly lower than that from WES (median: 39.508 vs. 49.773, P < 0.05). Our analysis reveals that as sequencing depth increases, the distribution of VAFs becomes more concentrated, evidenced by a significant reduction in the VAF MAD. Consequently, the MATH shows a significant decrease as sequencing depth increases (P < 0.05 for each comparison). While Shannon entropy displays a similar downward trend, it did not reach statistical significance in tissue samples when comparing WES and IMPACT panel sequencing (P = 0.260, S3 Fig). These findings suggest that depth-dependent VAF concentration exerts a stronger influence on MATH compared to Shannon entropy in high-coverage scenarios [23]. Last, to characterize the impact of sample size, we investigated the relationship between mutation data and ITH metrics using four tumor cohorts sequenced with the MSK-IMPACT panel and four same cancer types from TCGA cohorts (WES). We observed that both MATH and Shannon entropy exhibit minor fluctuations when the sample size is limited (n < 100) (S4 Fig). However, these fluctuations were consistent with the underlying shifts in tumor VAF distributions, indicating that the metrics remain stable and biologically coherent across diverse cohort sizes.
Case 3: Feasibility of plasma-based ITH assays.
To explore the potential of liquid biopsies in replacing tissue-based ITH assays, we analyzed 398 pancancer samples with paired tissue and plasma samples sequenced using identical NGS panels. While the correlation of MATH and Shannon entropy metrics between tissue and plasma was generally weak across most cancer types, distinct cohorts, including LIHC, CRC, RCC, and NSCLC showed significantly higher concordance (Fig 3K-L). This observed discordance suggests that the ITH profiles derived from tissue and plasma are likely driven by the distinct biological properties of the samples, specifically the localized spatial heterogeneity inherent in tissue biopsies versus the integrated systemic tumor burden captured by ctDNA, rather than technical biases or algorithmic limitations. To move beyond a theoretical concept and demonstrate the practical utility of this discordance, we introduced “Delta-ITH” (∆, defined as the tissue ITH score minus the plasma ITH score) as a novel dimension for understanding tumor evolution.
To rigorously validate this framework, we performed an in-depth clinical analysis on an independent cohort of 265 lung adenocarcinoma (LUAD, NM2022) patients who possessed paired tissue-plasma sequencing and complete clinical and follow-up data. Consistently, tissue-derived ITH metrics demonstrated a significant positive correlation with plasma-derived counterparts in this cohort (both MATH and Shannon entropy: R = 0.21, P < 0.001, S5A-B Fig). We next evaluated whether Delta-ITH correlated with basic demographic factors. While Delta-ITH did not significantly vary across age groups (<60 vs. ≥ 60 years), male patients displayed higher Delta-ITH values than female patients, which achieved statistical significance (∆-MATH: P = 0.092; ∆-Shannon entropy: P = 0.013, S5C-D Fig).
Crucially, we investigated the prognostic value of Delta-ITH by stratifying patients into a tissue-heterogeneity predominant group (Tissue-hetero, ∆ ≥ 0) and a circulating-heterogeneity predominant group (cfDNA-hetero, ∆ < 0). While no significant survival difference was observed using the MATH-based Delta-ITH stratification (Fig 4A-B), survival analysis revealed that based on Shannon entropy, the cfDNA-hetero group experienced significantly worse OS compared to the Tissue-hetero group (HR = 0.41, 95%CI: 0.22-0.79, Log-rank P = 0.006, Fig 4C). Multivariable Cox regression confirmed that this prognostic value was independent of standard clinical factors (P = 0.005, Fig 4D).
(A) Kaplan-Meier OS curves of LUAD patients stratified by MATH-based Delta-ITH; (B) Forest plot of multivariable Cox proportional hazards regression analysis evaluating the clinical independence of MATH-based Delta-ITH stratification alongside standard clinical covariates (age and gender); (C) Kaplan- Meier OS curves of LUAD patients stratified by Shannon entropy-based Delta-ITH; (D) Forest plot of multivariable Cox proportional hazards regression analysis evaluating the prognostic value of Shannon entropy-based Delta-ITH stratification.
Biologically, the poorer survival in the cfDNA-hetero population suggests that advanced tumor progression, often accompanied by accelerated cell apoptosis and necrosis, drives a higher systemic shedding of diverse subclonal DNA into the bloodstream. Consequently, plasma-derived Shannon entropy captures a more comprehensive, global tumor heterogeneity than a single localized tissue biopsy, aligning with established liquid biopsy paradigms where ctDNA dynamics reflect systemic tumor burden
Biologically, these findings suggest that in advanced-stage cancer patients, a higher volume of cfDNA is shed into the bloodstream. In this context, utilizing Shannon entropy on plasma samples captures global tumor heterogeneity more effectively than a single-site localized tissue biopsy. This observation aligns perfectly with methodological approaches adopted by previous hallmark liquid biopsy studies [18,20,24].
ITHindex supports the rapid benchmarking of novel algorithms. If a new algorithm is introduced, utilizing ITHindex to derive diverse ITH metric outcomes enables a rapid comparison of variances between the new algorithm and existing ones.
Discussion
ITHindex is the first interactive and easily accessible web-based tool designed for oncologic researchers to analyze the ITH of cancer patients, which is associated with the diverse biological mechanisms and treatment responses. ITHindex enables users to upload their own data and offers a variety of algorithms based on mutation, copy number variations, mRNA expression, protein abundance, and methylation data. By providing this integrated interface, ITHindex aims to bridge the gap between high-dimensional omics data and ITH quantification, ultimately providing an integrated resource for the cancer research community.
Although ITH has been shown to have a profound impact on tumor progression, immune evasion, and treatment response in various cancer types [25,26], assessing heterogeneity in human tumor samples remains a significant challenge [27]. Numerous methodologies have been developed to evaluate different aspects of tumor heterogeneity. However, current metrics are often empirical, challenging to interpret, and difficult to integrate into a unified framework, thereby limiting their reproducibility and adoption in clinical settings. In this study, we have presented several examples to illustrate how ITHindex can support ITH-related research. ITHindex enables the rapid acquisition of ITH scores from various algorithms, facilitating subsequent analysis. It is important to emphasize that ITHindex is designed as a research-enabling, exploratory platform. While our results demonstrate the platform’s potential in facilitating ITH investigations, the tool is not intended for clinical decision-making or diagnosis.
ITHindex can also assist in constructing NGS panel-based tissue ITH or blood ITH to assess ITH consistency, or in comparing new and existing algorithms to evaluate the performance attributes of the new algorithm.
Variations in ITH metric performance across sequencing platforms underscore the critical influence of sequencing depth and detection sensitivity on subclonal architecture quantification [28]. Our comparative analysis reveals that while deep sequencing (e.g., ctDNA panels at ~10,000x) captures a broader spectrum of low-frequency variants—thereby altering the VAF distribution profile—this does not uniformly translate into inflated MATH scores. This indicates that MATH is not merely a function of VAF dispersion (MAD) but is also heavily modulated by the median VAF of the sample [8]. In lower-depth WES data, the lack of sensitivity to the “long tail” of low-frequency subclonal variants shifts the median VAF upward and may lead to overestimation of heterogeneity due to sampling noise. In contrast, higher-depth platforms provide a more granular resolution of the subclonal landscape, resulting in a more robust and biologically representative ITH estimate.
Comparing tissue-derived and plasma-derived ITH metrics provides a unique window into tumor biology, reflecting the spatial and systemic dimensions of heterogeneity [29]. Our findings suggest that in the context of controlled sequencing depth, the weak correlation between tissue and plasma ITH is likely a biological signal rather than a technical artifact. Tissue samples, obtained via focal biopsy, reflect localized spatial heterogeneity within a specific tumor region. In contrast, ctDNA acts as a ‘liquid biopsy’ that integrates clonal information from the entire tumor burden, including primary lesions, multifocal tumors, and distant metastases [30]. Consequently, the difference between these two metrics—which we term “Delta-ITH”—may provide clinical value beyond individual measurements. By providing a unified platform to quantify ITH from both tissue and plasma, ITHindex empowers researchers to explore this Delta-ITH, offering a more nuanced understanding of how localized spatial heterogeneity contributes to systemic tumor evolution and treatment response. We explicitly note that the Case 3 verification cohort was derived from the LAVA database. This database comprises fully anonymized and de-identified data and has been utilized in several peer-reviewed studies [31–34]. Although this proprietary database offered unique deep-sequencing validation, we underscore that ITHindex remains a completely independent, vendor-neutral application designed to process standard Level 3 data from any user-provided clinical or public repository.
While ITHindex provides a unified and streamlined platform for the systematic evaluation of ITH, several limitations should be noted. First, the current framework is primarily tailored for bulk sequencing, containing only a single algorithm for single-cell data. Second, the metrics are inherently sensitive to upstream data preprocessing and technical covariates (e.g., tumor purity,sequencing depth and normalization methods) [28]. Because ITHindex focuses on Level 3 data to remain highly accessible for biomedical researchers without advanced coding skills, it does not integrate raw-data-based tools like epihet [35], or PhyloWG [10]. We therefore strongly recommend users adopt standardized workflows like GATK Best Practices [36,37] to ensure cross-study comparability. Third, while we comprehensively validated the cross--platform and cross-sample robustness of MATH and Shannon entropy-owing to their widespread adoption and minimal input barriers--the baseline technical dependencies for the remaining fifteen algorithms remain unestablished. This is primarily constrained by the extreme scarcity of public clinical cohorts that simultaneously provide matched multi-platform data for copy number, transcriptomic, proteomic, and epigenetic modalities. Users should thus interpret cross-platform comparisons for these extended metrics with caution. Additionally, the cross-omics application of certain algorithms -- such as applying sMPD to bulk transcriptomics or Shannon entropy to bulk proteomics -- remains exploratory and requires further methodological validation. Fourth, the Delta-ITH analysis serves as a preliminary, exploratory framework, as gene coverage and sequencing depths inherently vary between tissue and blood panels. Finally, while ITHindex is intended primarily as a research-enabling, exploratory tool for hypothesis generation, its current lack of integrated clinical phenotype data may limit its immediate utility in exploring complex clinical studies. Future upgrades will continuously incorporate newly published algorithms, accommodate larger multi-center cohorts with rigorous statistical adjustments, and bridge ITH metrics with comprehensive clinical metadata to enhance immediate translational utility.
The clinical translation of ITH as a robust biomarker faces significant challenges, particularly regarding the standardization of analytical pipelines and the reproducibility of results across different sequencing platforms. Furthermore, the biological significance of ITH is highly context-dependent, showing distinct evolutionary characteristics across various tumor types. In the realm of precision medicine, assessing ITH dynamics—especially through longitudinal liquid biopsies—offers critical value for predicting responses to immunotherapy and targeted therapy, as subclonal diversity often drives immune evasion and acquired resistance [38].
Moving forward, the integration of heterogeneity information across genomic, transcriptomic, and epigenomic layers represents a significant frontier for comprehensive tumor characterization. While ITHindex currently evaluates different data types independently, multi-omics integration can provide deeper insights into the evolutionary dynamics between genetic alterations and adaptive phenotypic changes. However, achieving this integration presents substantial technical challenges, including the need for standardized statistical frameworks to normalize diverse data scales and the requirement for matched multi-omics samples from the same biological regions. Future upgrades of ITHindex will explore these integrative methodologies, potentially utilizing advanced systems biology approaches to bridge the gap between distinct omics layers and provide a more holistic view of ITH.
In conclusion, ITHindex is an intuitive, registration-free, and easy-to-use analysis tool that can meet the needs of oncologic researchers in exploring the ITH based on multi-omics data using users’ own data. We believe that ITHindex will become a stable and reliable oncological tool with ongoing refinements.
Materials and methods
ITH algorithm identification and selection criteria
To ensure a systematic and transparent selection of the ITH quantification algorithms integrated into ITHindex, we performed a structured literature search across PubMed and Web of Science using the search string: (“intratumor heterogeneity” OR “intra-tumor heterogeneity” OR “ITH”) AND (“algorithm” OR “quantification” OR “metric” OR “measure”). The search was conducted from Jan 1990 to Dec 2025. The identification and screening process followed the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines, as illustrated in S1 Fig.
Inclusion criteria:
- Peer-Reviewed Validation: Algorithms must be supported by peer-reviewed literature, either as primary method development publications or through documented application in high-impact studies.
- Standardized Data Compatibility: Metrics must be compatible with Level 3 “processed” omics data (e.g., mutation, expression matrices, methylation beta values), ensuring accessibility for researchers without high-performance computing infrastructure for raw data processing.
- Algorithmic Transparency & Reproducibility: Algorithms must provide sufficient methodological documentation, including detailed step-by-step mathematical formulations, pseudocode, or accessible open-source implementations (e.g., R, Python, MATLAB), allowing for robust integration into our automated pipeline.
Exclusion Criteria:
- Lack of Validation: Metrics or algorithms lacking documentation in peer-reviewed journals.
- Requirement for Raw Data: Algorithms strictly requiring raw data formats (e.g., FASTQ, BAM) without options for processing Level 3 data.
- Methodological Opacity: Algorithms with proprietary implementations, inaccessible source code, or insufficient documentation.
Pan-cancer data collection
We gathered human tumor-related data from three major publicly available databases: The Cancer Genome Atlas (TCGA), Gene Expression Omnibus (GEO) (https://www.ncbi.nlm.nih.gov/geo/), and cBioPortal (https://www.cbioportal.org/). Specifically, we obtained clinical information, WES data, copy number (CN) segments, RNA sequencing (RNAseq), and DNA methylation 27K microarray data for 32 tumor types from the TCGA database through cBioPortal. Additionally, we collected bulk expression profiles from four tumor types with multi-region sampling from the GEO database, including lung cancer (GSE33532) [39], breast cancer (GSE23593) [40], liver cancer (GSE136711) [41], and cervical cancer (GSE5787) [42]. Among these, GSE136711 provides RNAseq data, while the other three datasets offer mRNA microarray data. Additionally, five cohorts were retrieved from cBioPortal: three MSK-IMPACT targeted panel cohorts, including CRC (Gastro2020) [43], HCC (CCR2024) [44], and BRCA (CC2018) [45]; and two LUAD cohorts — LUAD (NM2022) and LUAD (CC2023) [46,47] — which encompass multi-platform sequencing data, such as tissue WES (~150x), tissue panels (MSK-IMPACT, ~ 800x), and plasma panels (~10,000x). Finally, we included in-house anonymization data from 398 patients who underwent next-generation sequencing (NGS) of a 520-gene panel (~10000x) on tumor DNA derived from both plasma and tissue samples, encompassing 16 types of solid tumors from the Burning Rock LAVA database (Burning Rock Biotech, Guangzhou, China). Altogether, 11,242 TCGA pan-cancer samples, 4,904 cBioPortal pan-cancer samples, and 398 in-house cases were included in this study. Detailed cohort information is provided in S2 Table.
Application implementation
ITHindex was developed using the Shiny framework, enabling interactive data analysis and visualization with R. It leverages the tidyverse family of R packages (e.g., dplyr, purrr, and tibble) for internal tabular data transformations, the DT package for interactive data tables, and the ggplot2 and plotly packages for visualizing the ITH index. Static plots are created using ggplot2, while interactive plots are generated with the plotly package.
ITHindex can be deployed on the web server provided by shinyapps.io or on a local virtual machine using the ShinyProxy framework (https://www.shinyproxy.io), which organizes and manages Shiny application instances using Docker containers based on user demand. As a result, installation of R and RStudio on the user’s device is not required. The web-based tool is accessible via the following links: https://shinyapps.brbiotech.com/app/ithindex or https://xwen.shinyapps.io/ithindex/. Users can access the analysis tool through any standard browser, including Firefox, Chrome, Safari, or Microsoft Edge.
ITHindex is hosted on a server equipped with an 8-core CPU and 32 GB RAM to ensure robust computational performance and scalability. For a standard clinical cohort (e.g., 200 samples), the platform typically completes ITH quantification across multiple algorithms within 10 seconds, depending on the computational complexity of the selected metrics. To maintain infrastructure stability during user access, ITHindex incorporates proactive data validation, including format checks (e.g., file extension, column naming, and data type consistency) to identify malformed inputs before analysis. Furthermore, the web server currently supports a maximum upload file size of 50 MB per session, which accommodates the majority of processed “Level 3” multi-omics datasets. For larger pan-cancer datasets, we recommend that users upload data in smaller batches to adhere to file size constraints while maintaining optimal processing speeds.
To ensure long-term sustainability, ITHindex utilizes a structured maintenance strategy, with source code publicly hosted on GitHub under semantic versioning for transparent update tracking. Infrastructure stability is guaranteed by institutional support, which provides dedicated server resources for continued accessibility. We are committed to maintaining the web deployment for a minimum of five years post-publication, supported by an email-driven issue-tracking mechanism to facilitate ongoing user feedback and collaborative enhancement.
Supporting information
S1 Fig. PRISMA flow diagram for the literature search.
https://doi.org/10.1371/journal.pdig.0001654.s001
(TIF)
S2 Fig. ITH metrics and relationships.
A. Collected ITH metrics, organized by type of heterogeneity, mathematical theory, and data type. B-C. The correlation between sMPD and diversity score for the same data type in HCC (B) and Glioma cancers (C).
https://doi.org/10.1371/journal.pdig.0001654.s002
(TIFF)
S3 Fig. ITH metric divergence across varied sequencing platforms.
A. Comparative analysis of MATH and Shannon entropy between MSK-IMPACT and WES in paired LUAD patients; B. Comparative analysis of MATH and Shannon entropy between cfDNA panel and MSK-IMPACT in paired LUAD patients; C. Comparative analysis of MATH and Shannon entropy between cfDNA panel and WES in paired LUAD patients.
https://doi.org/10.1371/journal.pdig.0001654.s003
(TIFF)
S4 Fig. The trends of ITH metrics in relation to sample size across multiple cancer cohorts.
A. CRC; B. HCC; C. BRCA; D. LUAD.
https://doi.org/10.1371/journal.pdig.0001654.s004
(TIFF)
S5 Fig. Comparison of tissue and plasma ITH metrics and clinical association of Delta-ITH in LUAD patients.
(A, B) Correlation analysis of (A) MATH and (B) Shannon entropy between paired tissue and plasma samples in LUAD patients. (C, D) Boxplots showing the association of (C) MATH-based and (D) Shannon entropy-based Delta-ITH with clinical characteristics (age and gender).
https://doi.org/10.1371/journal.pdig.0001654.s005
(TIFF)
S1 Table. Collected intra-tumor heterogeneity metrics, organized by type of heterogeneity, mathematical theory, and data type.
https://doi.org/10.1371/journal.pdig.0001654.s006
(DOCX)
S2 Table. Detailed information of cohort used in this study.
https://doi.org/10.1371/journal.pdig.0001654.s007
(DOCX)
Acknowledgments
The icons used in Fig 1 were sourced from Flaticon.com. These icons were created by Victoruler, Pikepicture, Riajulislam, Freepik, and Vichanon Chaimsuk, and are used under the Free License with attribution.
References
- 1. Ramón Y Cajal S, Sesé M, Capdevila C, Aasen T, De Mattos-Arruda L, Diaz-Cano SJ, et al. Clinical implications of intratumor heterogeneity: challenges and opportunities. J Mol Med (Berl). 2020;98(2):161–77. pmid:31970428
- 2. Gao Z, Jiang A, Li Z, Zhu L, Mou W, Shen W, et al. Heterogeneity of intratumoral microbiota within the tumor microenvironment and relationship to tumor development. Med Research. 2025;1(1):32–61.
- 3. Lin A, Xiong M, Jiang A, Huang L, Wong HZH, Feng S, et al. The microbiome in cancer. Imeta. 2025;4(5):e70070. pmid:41112042
- 4. Chen L, Lin A, Tang B, Li K, Qian Y, Liu X. The future of cancer therapy: nanomaterials and tumor microenvironment. iMetaMed. 2025;1:e70007.
- 5. Caswell DR, Swanton C. The role of tumour heterogeneity and clonal cooperativity in metastasis, immune evasion and clinical outcome. BMC Med. 2017;15(1):133. pmid:28716075
- 6. Jamal-Hanjani M, Quezada SA, Larkin J, Swanton C. Translational implications of tumor heterogeneity. Clin Cancer Res. 2015;21(6):1258–66. pmid:25770293
- 7. Jaccard P. The distribution of the flora in the alpine zone. New Phytologist. 1912;11(2):37–50.
- 8. Mroz EA, Rocco JW. MATH, a novel measure of intratumor genetic heterogeneity, is high in poor-outcome classes of head and neck squamous cell carcinoma. Oral Oncol. 2013;49(3):211–5. pmid:23079694
- 9. Andor N, Harness JV, Müller S, Mewes HW, Petritsch C. EXPANDS: expanding ploidy and allele frequency on nested subpopulations. Bioinformatics. 2014;30(1):50–60. pmid:24177718
- 10. Deshwar AG, Vembu S, Yung CK, Jang GH, Stein L, Morris Q. PhyloWGS: reconstructing subclonal composition and evolution from whole-genome sequencing of tumors. Genome Biol. 2015;16(1):35. pmid:25786235
- 11. McGranahan N, Favero F, de Bruin EC, Birkbak NJ, Szallasi Z, Swanton C. Clonal status of actionable driver events and the timing of mutational processes in cancer evolution. Sci Transl Med. 2015;7(283):283ra54. pmid:25877892
- 12. Dixon P. VEGAN, a package of R functions for community ecology. J Vegetation Science. 2003;14(6):927–30.
- 13. Grendár M, Martínek P, Loderer D, Ondič O. CNHplus: the chromosomal copy number heterogeneity which respects biological constraints. bioRxiv. 2022;:2022.09.30.510279.
- 14. Park Y, Lim S, Nam J-W, Kim S. Measuring intratumor heterogeneity by network entropy using RNA-seq data. Sci Rep. 2016;6:37767. pmid:27883053
- 15. Song D, Wang X. DEPTH2: an mRNA-based algorithm to evaluate intratumor heterogeneity without reference to normal controls. J Transl Med. 2022;20(1):150. pmid:35365157
- 16. Zhang JD, Hatje K, Sturm G, Broger C, Ebeling M, Burtin M, et al. Detect tissue heterogeneity in gene expression data with BioQC. BMC Genomics. 2017;18(1):277. pmid:28376718
- 17. Kashyap A, Rapsomaniki MA, Barros V, Fomitcheva-Khartchenko A, Martinelli AL, Rodriguez AF, et al. Quantification of tumor heterogeneity: from data acquisition to metric generation. Trends Biotechnol. 2022;40(6):647–76. pmid:34972597
- 18. Fan Y, Liu Y, Wang L, Cai Y, Cao W, Sun W, et al. bITH, a blood-based metric of intratumor heterogeneity, is associated with clinical response to immune checkpoint blockade in non-small cell lung cancer. EBioMedicine. 2023;91:104564. pmid:37094467
- 19. Brooks FJ. On some misconceptions about tumor heterogeneity quantification. Eur J Nucl Med Mol Imaging. 2013;40(8):1292–4. pmid:23632962
- 20. Zhou J, Bao M, Gao G, Cai Y, Wu L, Lei L, et al. Increased blood-based intratumor heterogeneity (bITH) is associated with unfavorable outcomes of immune checkpoint inhibitors plus chemotherapy in non-small cell lung cancer. BMC Med. 2022;20(1):256. pmid:35902848
- 21. Yaung SJ, Ju C, Gattam S, Nicholas A, Sommer N, Bendell JC, et al. Plasma-based measurements of tumor heterogeneity correlate with clinical outcomes in metastatic colorectal cancer. Cancers (Basel). 2022;14(9):2240. pmid:35565368
- 22. Ma D, Jiang YZ, Liu XY, Liu YR, Shao ZM. Clinical and molecular relevance of mutant-allele tumor heterogeneity in breast cancer. Breast Cancer Research and Treatment. 2017;162:39–48.
- 23. Noorbakhsh J, Kim H, Namburi S, Chuang JH. Distribution-based measures of tumor heterogeneity are sensitive to mutation calling and lack strong clinical predictive power. Sci Rep. 2018;8(1):11445. pmid:30061557
- 24. Xu J, Wan R, Cai Y, Cai S, Wu L, Li B, et al. Circulating tumor DNA-based stratification strategy for chemotherapy plus PD-1 inhibitor in advanced non-small-cell lung cancer. Cancer Cell. 2024;42(9):1598-1613.e4. pmid:39255777
- 25. McGranahan N, Swanton C. Biological and therapeutic impact of intratumor heterogeneity in cancer evolution. Cancer Cell. 2015;27(1):15–26. pmid:25584892
- 26. Bedard PL, Hansen AR, Ratain MJ, Siu LL. Tumour heterogeneity in the clinic. Nature. 2013;501(7467):355–64. pmid:24048068
- 27. Marusyk A, Janiszewska M, Polyak K. Intratumor heterogeneity: the rosetta stone of therapy resistance. Cancer Cell. 2020;37(4):471–84. pmid:32289271
- 28. Abécassis J, Hamy A-S, Laurent C, Sadacca B, Bonsang-Kitzis H, Reyal F, et al. Assessing reliability of intra-tumor heterogeneity estimates from single sample whole exome sequencing data. PLoS One. 2019;14(11):e0224143. pmid:31697689
- 29. Gilson P, Merlin J-L, Harlé A. Deciphering tumour heterogeneity: from tissue to liquid biopsy. Cancers (Basel). 2022;14(6):1384. pmid:35326534
- 30. Lone SN, Nisar S, Masoodi T, Singh M, Rizwan A, Hashem S, et al. Liquid biopsy: a step closer to transform diagnosis, prognosis and future of cancer treatments. Mol Cancer. 2022;21(1):79. pmid:35303879
- 31. Jia Z, Xing J, Li J, Wang W, Wang Y, Song Y, et al. HER2 transmembrane domain mutation: comprehensive characteristics and real-world evidence of treatment response in Chinese lung adenocarcinoma. Transl Lung Cancer Res. 2021;10(3):1383–96. pmid:33889517
- 32. Sun D, Wu W, Wang L, Qu J, Han Q, Wang H, et al. Identification of MET fusions as novel therapeutic targets sensitive to MET inhibitors in lung cancer. J Transl Med. 2023;21(1):150. pmid:36829199
- 33. Xiang C, Lv X, Chen K, Guo L, Zhao R, Teng H, et al. Unraveling the significance of MET focal amplification in lung cancer: integrative NGS, FISH, and IHC investigation. Mod Pathol. 2024;37(4):100451. pmid:38369190
- 34. Zhang X, Liu N, Song S, Zhang C, Li T, Song W, et al. Next-generation sequencing identifies potential mechanisms of primary resistance to trastuzumab-containing therapies in HER2-positive gastric cancer: analysis based on real-world data and online databases. Hum Vaccin Immunother. 2025;21(1):2555502. pmid:40922695
- 35. Chen X, Ashoor H, Musich R, Wang J, Zhang M, Zhang C, et al. epihet for intra-tumoral epigenetic heterogeneity analysis and visualization. Sci Rep. 2021;11(1):376. pmid:33432081
- 36. Koboldt DC. Best practices for variant calling in clinical sequencing. Genome Med. 2020;12(1):91. pmid:33106175
- 37. Krusche P, Trigg L, Boutros PC, Mason CE, De La Vega FM, Moore BL, et al. Best practices for benchmarking germline small-variant calls in human genomes. Nat Biotechnol. 2019;37(5):555–60. pmid:30858580
- 38. Goyette M-A, Lipsyc-Sharf M, Polyak K. Clinical and translational relevance of intratumor heterogeneity. Trends Cancer. 2023;9(9):726–37. pmid:37248149
- 39.
Meister M, Belousov A, Xu EC, Albert PS. Intra-tumor heterogeneity of gene expression profiles in early stage non-small cell lung cancer.
- 40. Barry WT, Kernagis DN, Dressman HK, Griffis RJ, Hunter JD, Olson JA, et al. Intratumor heterogeneity and precision of microarray-based predictors of breast cancer biology and clinical outcome. J Clin Oncol. 2010;28(13):2198–206. pmid:20368555
- 41. Cherry TJ, Yang MG, Harmin DA, Tao P, Timms AE, Bauwens M, et al. Mapping the cis-regulatory architecture of the human retina reveals noncoding genetic variation in disease. Proc Natl Acad Sci U S A. 2020;117(16):9001–12. pmid:32265282
- 42. Bachtiary B, Boutros PC, Pintilie M, Shi W, Bastianutto C, Li J-H, et al. Gene expression profiling in cervical cancer: an exploration of intratumor heterogeneity. Clin Cancer Res. 2006;12(19):5632–40. pmid:17020965
- 43. Mondaca S, Walch H, Nandakumar S, Chatila WK, Schultz N, Yaeger R. Specific mutations in APC, but not alterations in DNA damage response, associate with outcomes of patients with metastatic colorectal cancer. Gastroenterology. 2020;159(5):1975-1978.e4. pmid:32730818
- 44. Song Y, Boerner T, Drill E, Shin P, Kumar S, Sigel C, et al. A novel approach to quantify heterogeneity of intrahepatic cholangiocarcinoma: the hidden-genome classifier. Clin Cancer Res. 2024;30(16):3499–511. pmid:38864854
- 45. Razavi P, Chang MT, Xu G, Bandlamudi C, Ross DS, Vasan N, et al. The genomic landscape of endocrine-resistant advanced breast cancers. Cancer Cell. 2018;34(3):427-438.e6. pmid:30205045
- 46. Jee J, Lebow ES, Yeh R, Das JP, Namakydoust A, Paik PK, et al. Overall survival with circulating tumor DNA-guided therapy in advanced non-small-cell lung cancer. Nat Med. 2022;28(11):2353–63. pmid:36357680
- 47. Lengel HB, Mastrogiacomo B, Connolly JG, Tan KS, Liu Y, Fick CN, et al. Genomic mapping of metastatic organotropism in lung adenocarcinoma. Cancer Cell. 2023;41(5):970-985.e3. pmid:37084736