Home Journals Articles Updates About
CN
image
收藏切换
Artificial Intelligence-Powered Insights into Polyclonality and Tumor Evolution
收藏切换
PDF
Hong Zhao1, 2, *, Trey Ideker3, Stephen T. C. Wong1, 4, *
Research. Vol 8 Article ID 0765
Less
收藏切换
Research. Vol 8 Article ID 0765
• Perspective •
Artificial Intelligence-Powered Insights into Polyclonality and Tumor Evolution
Full
Hong Zhao1, 2, *, Trey Ideker3, Stephen T. C. Wong1, 4, *
Affiliations
  • 1Department of Systems Medicine and Bioengineering, Houston Methodist Neal Cancer Center, Houston Methodist Hospital, Houston, TX 77030, USA.
  • 2Department of Medicine, Weill Cornell Medicine, New York, NY 10065, USA.
  • 3Department of Medicine, University of California San Diego, La Jolla, CA 92093, USA.
  • 4Departments of Radiology, Pathology, and Laboratory Medicine, Weill Cornell Medicine, New York, NY 10065, USA.
Published: 2025-07-20 doi: 10.34133/research.0765
Outline
收藏切换

Recent studies have revealed that polyclonality—where multiple distinct subclones cooperate during early tumor development—is a critical feature of tumor evolution, as demonstrated by Sadien et al. and Lu et al. in Nature (October 2024). These findings show that early polyclonal interactions can overcome fitness barriers, ultimately transitioning to monoclonality as dominant clones emerge. Understanding and targeting these interclonal dynamics offers new therapeutic opportunities. In this perspective, we outline how computational modeling and artificial intelligence (AI) tools can provide deeper insights into tumor polyclonality and identify actionable therapeutic strategies. By applying ligand–receptor interaction analysis, clonal trajectory reconstruction, network and pathway modeling, and spatial analysis, researchers can prioritize communication hubs, evolutionary bottlenecks, and microenvironmental niches that sustain tumor progression. These approaches, when integrated with experimental validation, offer a translational pathway from foundational discoveries to personalized cancer treatments aimed at disrupting cooperative subclonal ecosystems and preventing malignant progression.We commend the recent Nature publications, “Polyclonality overcomes fitness barriers in Apc-driven tumorigenesis” by Sadien et al. [1] and “Polyclonal-to-monoclonal transition in colorectal precancerous evolution” by Lu et al. [2], both featured on 2024 October 30. These groundbreaking studies employed distinct lineage tracing methods to investigate the origins and evolutionary dynamics of colorectal and intestinal tumorigenesis. Despite their different approaches, both studies reached convergent conclusions: Polyclonality plays a pivotal role in the early stages of tumor development, providing critical insights into how diverse cellular populations collaborate to overcome fitness barriers and drive tumor progression.

Hong Zhao, Trey Ideker, Stephen T. C. Wong. Artificial Intelligence-Powered Insights into Polyclonality and Tumor Evolution[J]. Research, 2025 , 8 (7) : 0765 - . DOI: 10.34133/research.0765
Full Text
Less
Clarifying Definitions: Polyclonality and Monoclonality in Tumor Evolution
Less
In this perspective, we define polyclonality as the presence of multiple independent ancestral lineages contributing to the formation of a single precancerous lesion, each arising separately and harboring distinct sets of somatic mutations. This is consistent with the definition used by Lu et al. [2], who demonstrated through base editor-enabled DNA barcoding that early-stage colorectal lesions often originate from multiple distinct clones. These lineages coexist and interact within the same tissue microenvironment without any single clone initially dominating. The transition to monoclonality refers to the evolutionary bottleneck observed during tumor progression, where competitive pressures—such as selective growth advantages from driver mutations—lead to the outgrowth of one dominant lineage, resulting in a monoclonal tumor mass derived from a single ancestral clone. As shown by Lu et al., this transition is associated with diminished interclonal interactions and a collapse of clonal diversity over time.
Introduction of the 2 Studies
Less
The Sadien study [1] used multicolor lineage tracing with chemical mutagenesis to demonstrate that many intestinal tumors have a polyclonal origin, consisting of subclones with distinct APC mutations. These diverse subclones are influenced by differences in KRAS and MYC signaling and highlight the role of interclonal interactions in tumor progression. In parallel, Lu et al. [2] employed a base editor-enabled DNA barcoding system to map single-cell phylogenies in mouse models of colorectal tumorigenesis. Their analysis of over 260,000 single cells found that colorectal precancerous lesions often arise from multiple lineages. Early polyclonal lesions displayed extensive intercellular interactions, which were significantly diminished during the transition to monoclonality.
These studies underscore the importance of polyclonality and interclonal interactions in early tumor formation. Intratumor heterogeneity, resulting from clonal evolution and branching, has been widely observed across many cancers, including breast, lung, prostate tumors, and glioblastoma [3,4]. Multiple clones contribute to tumor growth, therapy resistance, and metastasis, emphasizing the need to understand interclonal interactions. Targeting the cooperation or competition between clones also offers opportunities to control tumor progression. Understanding and overcoming these complexities are also crucial in developing effective, long-term cancer therapies for improving cancer treatment outcomes. All of these areas currently rely critically on strong computational analysis tools, and they suggest opportunities for sustained computational method development.
We propose that integrating data from studies like those by Sadien et al. and Lu et al. with advanced computational modeling and artificial intelligence (AI) can yield powerful insights for translational research, from early detection and targeted interventions to personalized treatment plans, ultimately aiming to improve patient outcomes (Figure).
Modeling ligand–receptor interactions
In the context of modeling ligand–receptor interactions, tools such as CellChat [5] can analyze single-cell RNA sequencing (scRNA-seq) data to identify ligand–receptor interactions between subclones. For example, Lu et al. [2] demonstrated 14 enriched ligand–receptor interactions in early polyclonal lesions, involving extracellular matrix (ECM) organization and cell adhesion—processes critical for maintaining clonal cooperation. CCCExplorer [6,7] complements this by mapping how these interactions funnel into downstream signaling pathways. As demonstrated by Lu et al. [2], mutations in genes such as BCL9L, SOX9, TCF7L2, and CTNNB1 diminish interclonal communication during the transition to monoclonality. Tools like CCCExplorer can link these mutations to specific ligand–receptor pathways, revealing how signaling disruptions promote the loss of clonal cooperation.
AI enhances these tools by enabling the prediction and refinement of ligand–receptor interactions. Machine learning algorithms, such as random forests or neural networks, can integrate scRNA-seq data to predict novel ligand–receptor pairs that drive interclonal communication [8]. Autoencoders, a class of unsupervised deep learning models, can be applied to single-cell transcriptomic data to extract subclone-specific latent features. By compressing and reconstructing gene expression profiles, autoencoders can identify signaling modules preferentially activated within distinct subclones. Analysis of these modules can reveal how specific subclones contribute to tumor growth, immune evasion, or therapy resistance [9]. Bayesian network models can be used to infer probabilistic causal relationships between somatic mutations and downstream pathway activation within individual clones. While less established, reinforcement learning could be adapted to simulate clonal competition dynamics and optimize therapeutic strategies that disrupt critical interclonal interactions, using fitness reduction or loss of cooperative signaling as a reward function [10].
Modeling ligand–receptor interactions enables the identification of key interclonal communication pathways that sustain early tumor polyclonality. Therapeutically, targeting these pathways—for example, disrupting ECM remodeling or adhesion signals that maintain clonal cooperation—could selectively destabilize cooperative subclonal ecosystems, impairing tumor progression at its earliest stages.
Reconstructing clonal evolution with trajectory tools
The reconstruction of clonal evolution, as explored by both Sadien et al. and Lu et al., can also benefit from AI. Tools such as Monocle [11] trace evolutionary trajectories of cell populations, linking specific mutations to changes in clonal behavior. Lu et al. highlighted how mutations in BCL9L and CTNNB1 drive clonal dominance during monoclonal transitions. Deep variational autoencoders (VAEs) can capture branching events in clonal evolution [12], while natural language processing models can annotate mutations like APC or KRAS, contextualizing their functional relevance. Recurrent neural networks and time-series modeling can forecast when subclones gain dominance [13], providing insights into how clonal cooperation diminishes over time.
Reconstructing clonal evolution trajectories allows researchers to map when specific subclones acquire dominance-driving features, such as key mutations or pathway activations. Targeting these critical events therapeutically could delay or prevent the collapse into monoclonality, potentially maintaining a more indolent polyclonal state or rendering tumors more susceptible to therapy.
Network and pathway analysis
Network and pathway analysis further elucidate how clonal interactions shape tumor evolution. Sadien et al. emphasized diverse signaling profiles among APC-mutant subclones, while Lu et al. linked specific mutations to disrupted ECM and adhesion-related pathways. Tools like Cytoscape [14] can visualize these networks, showing how signaling disruptions lead to clonal competition. Graph neural networks expand on this by identifying cooperative signaling modules [15], while machine learning simulators model how perturbations in nodes like MYC affect interclonal interactions [16]. Deep VAEs, as demonstrated by Geleta et al. [12], can compress complex genotype data into lower-dimensional latent feature spaces while preserving meaningful biological structure. In the context of tumor evolution, similar strategies could be adapted to compress single-cell signaling profiles, where subsequent clustering or anomaly detection analyses might reveal emergent patterns of clonal cooperation or dominance. While promising, these applications remain exploratory and require further validation in cancer models.
Network and pathway analysis can reveal central nodes critical for sustaining interclonal cooperation, such as MYC-driven proliferative signaling or CTNNB1-mediated adhesion. Therapeutic targeting of these central hubs could dismantle the cooperative network architecture, selectively impairing tumor growth and clonal expansion.
Spatial analysis of clonal distribution
Spatial analysis of clonal distribution, highlighted by Sadien et al. and Lu et al., reveals how subclones occupy distinct niches and interact within the tumor microenvironment. Spatial transcriptomic tools such as SpatialDE [17] map how mutations influence spatial interactions between clones. Lu et al. observed a reduction in interclonal proximity and communication as lesions transitioned to monoclonality. Convolutional neural networks (CNNs) can analyze histological images to track clonal clustering or dispersal [18], while spatial graph convolutional networks quantify cell–cell interactions sustaining polyclonality [19–21]. Agent-based modeling integrates morphological data with AI decision engines, simulating how clones compete or cooperate in real-world conditions [16].
Integrating AI-driven computational analyses with single-cell and spatial omics data offers not only a deeper understanding of early tumor polyclonality but also a practical framework for therapeutic discovery. By identifying interclonal signaling dependencies, reconstructing evolutionary bottlenecks, and mapping cooperative spatial niches [22], these tools can prioritize actionable targets for disrupting subclonal cooperation, delaying clonal dominance, and personalizing treatment strategies for heterogeneous tumors.
Spatial analysis allows mapping of where cooperative or dominant subclones reside within the tumor microenvironment. Therapeutically, spatial targeting—such as site-specific delivery of immune modulators or pathway inhibitors—could be designed to selectively disrupt localized niches supporting polyclonal cooperation or emerging clonal dominance.
Critical analysis of AI and computational tools
Although computational modeling and AI-driven tools have opened new avenues for studying tumor polyclonality, several technical and practical limitations remain. The predictive power of these models heavily relies on the quality, size, and diversity of input datasets. However, many publicly available scRNA-seq and spatial transcriptomic datasets are relatively small, suffer from sampling biases (e.g., overrepresentation of specific tumor types or stages), and often lack paired clinical annotations such as treatment histories or patient outcomes. These gaps limit the models’ ability to generalize across patient populations.
Furthermore, most AI models, especially deep learning frameworks like CNNs and autoencoders, operate as “black boxes”, producing predictions without clear mechanistic explanations. This lack of interpretability complicates biological validation and limits clinical translation. Overfitting is another major concern, particularly when training datasets are small or unbalanced, leading models to capture noise instead of true biological signals.
Algorithmic biases can also emerge when training data reflect historical inequities or technological artifacts, potentially reinforcing errors rather than correcting them. These issues are especially critical when inferring interclonal communications or reconstructing evolutionary trajectories, where biological nuance matters.
To overcome these challenges, rigorous model validation using independent experimental datasets, cross-cohort comparisons, and standardized benchmarking practices are essential. Expanding multi-institutional data sharing efforts and integrating multi-modal data (e.g., genomics, transcriptomics, proteomics, and imaging) can also improve the robustness and clinical relevance of AI-driven insights into tumor evolution.
Funding
Less
  • Basic Research Laboratory(NIH R01 5CA238727)
  • Basic Research Laboratory(NIH R01 5CA238727)
  • Basic Research Laboratory(NIH U01CA252553)
References
Less
1.
Sadien ID, Adler S, Mehmed S, Bailey S, Sawle A, Couturier DL, Eldridge M, Adams DJ, Kemp R, Lourenço FC, et al.Polyclonality overcomes fitness barriers in Apc-driven tumorigenesis. Nature. 2024;634(8036):1196–1203.
2.
Lu Z, Mo S, Xie D, Zhai X, Deng S, Zhou K, Wang K, Kang X, Zhang H, Tong J, et al.Polyclonal-to-monoclonal transition in colorectal precancerous evolution. Nature. 2024;636(8041):233–240.
3.
Dentro SC, Leshchiner I, Haase K, Tarabichi M, Wintersinger J, Deshwar AG, Yu K, Rubanova Y, Macintyre G, Demeulemeester J, et al.Characterizing genetic intra-tumor heterogeneity across 2,658 human cancer genomes. Cell. 2021;184(8):2239–2254.e39.
4.
Gerlinger M, Rowan AJ, Horswell S, Math M, Larkin J, Endesfelder D, Gronroos E, Martinez P, Matthews N, Stewart A, et al.Intratumor heterogeneity and branched evolution revealed by multiregion sequencing. N Engl J Med. 2012;366(10):883–892.
5.
Jin S, Guerrero-Juarez CF, Zhang L, Chang I, Ramos R, Kuan CH, Myung P, Plikus MV, Nie Q. Inference and analysis of cell-cell communication using CellChat. Nat Commun. 2021;12(1):1088.
6.
Choi H, Sheng J, Gao D, Li F, Durrans A, Ryu S, Lee SB, Narula N, Rafii S, Elemento O, et al.Transcriptome analysis of individual stromal cell populations identifies stroma-tumor crosstalk in mouse lung cancer model. Cell Rep. 2015;10(7):1187–1201.
7.
Yeung TL, Sheng J, Leung CS, Li F, Kim J, Ho SY, Matzuk MM, Lu KH, Wong STC, Mok SC. Systematic identification of druggable epithelial-stromal crosstalk signaling networks in ovarian cancer. J Natl Cancer Inst. 2019;111(3):272–282.
8.
Lu M, Sha Y, Silva TC, Colaprico A, Sun X, Ban Y, Wang L, Lehmann BD, Chen XS. LR hunting: A random forest based cell-cell interaction discovery method for single-cell gene expression data. Front Genet. 2021;12:Article 708835.
9.
Caravagna G, Heide T, Williams MJ, Zapata L, Nichol D, Chkhaidze K, Cross W, Cresswell GD, Werner B, Acar A, et al.Subclonal reconstruction of tumors by using machine learning and population genetics. Nat Genet. 2020;52(9):898–907.
10.
Sun Z, Chung D, Neelon B, Millar-Wilson A, Ethier SP, Xiao F, Zheng Y, Wallace K, Hardiman G. A Bayesian framework for pathway-guided identification of cancer subgroups by integrating multiple types of genomic data. Stat Med. 2023;42(28):5266–5284.
11.
Trapnell C, Cacchiarelli D, Grimsby J, Pokharel P, Li S, Morse M, Lennon NJ, Livak KJ, Mikkelsen TS, Rinn JL. The dynamics and regulators of cell fate decisions are revealed by pseudotemporal ordering of single cells. Nat Biotechnol. 2014;32(4):381–386.
12.
Geleta M, Montserrat DM, Giro-i-Nieto X, Ioannidis AG. Deep variational autoencoders for population genetics. bioRxiv. 2023. https://doi.org/10.1101/2023.09.27.558320
13.
Hewamalage H, Bergmeir C, Bandara K. Recurrent neural networks for time series forecasting: Current status and future directions. Int J Forecast. 2021;37(1):388–427.
14.
Shannon P, Markiel A, Ozier O, Baliga NS, Wang JT, Ramage D, Amin N, Schwikowski B, Ideker T. Cytoscape: A software environment for integrated models of biomolecular interaction networks. Genome Res. 2003;13(11):2498–2504.
15.
Xingyi Li JX, Li J, Gu J, Shang X. Towards simplified graph neural networks for identifying cancer driver genes in heterophilic networks. Brief Bioinform. 2025;3(26):Article bbae691.
16.
Cogno N, Axenie C, Bauer R, Vavourakis V. Agent-based modeling in cancer biomedicine: Applications and tools for calibration and validation. Cancer Biol Ther. 2024;25(1):2344600.
17.
Svensson V, Teichmann SA, Stegle O. SpatialDE: Identification of spatially variable genes. Nat Methods. 2018;15(5):343–346.
18.
Sharkas M, Attallah O. Color-CADx: A deep learning approach for colorectal cancer classification through triple convolutional neural networks and discrete cosine transform. Sci Rep. 2024;14(1):6914.
19.
Li Y, Luo Y. STdGCN: Spatial transcriptomic cell-type deconvolution using graph convolutional networks. Genome Biol. 2024;25(1):206.
20.
Wang H, Zhao J, Nie Q, Zheng C, Sun X. Dissecting spatiotemporal structures in spatial transcriptomics via diffusion-based adversarial learning. Research. 2024;7:0390.
21.
Li Y, Lu Y, Kang C, Li P, Chen L. Revealing tissue heterogeneity and spatial dark genes from spatially resolved transcriptomics by multiview graph networks. Research. 2023;6:0228.
22.
Bhinder B, Gilvary C, Madhukar NS, Elemento O. Artificial intelligence in cancer research and precision medicine. Cancer Discov. 2021;11(4):900–915.
Appendix
Less
Year 2025 volume 8 Issue 7
PDF
46
26
Cite this Article
BibTeX
Article Info
doi: 10.34133/research.0765
  • Receive Date：2025-01-22
  • Online Date：2026-10-30
  • Published：2025-07-20
Article Data
Affiliations
History
  • Received：2025-01-22
  • Revised：2025-04-29
  • Accepted：2025-06-12
Funding
Basic Research Laboratory(NIH R01 5CA238727)
Basic Research Laboratory(NIH R01 5CA238727)
Basic Research Laboratory(NIH U01CA252553)
Affiliations
    1Department of Systems Medicine and Bioengineering, Houston Methodist Neal Cancer Center, Houston Methodist Hospital, Houston, TX 77030, USA.
    2Department of Medicine, Weill Cornell Medicine, New York, NY 10065, USA.
    3Department of Medicine, University of California San Diego, La Jolla, CA 92093, USA.
    4Departments of Radiology, Pathology, and Laboratory Medicine, Weill Cornell Medicine, New York, NY 10065, USA.

Corresponding:

* Address correspondence to: (H.Z.);
(S.W.)
References
Share
https://castjournals.cast.org.cn/joweb/research/EN/10.34133/research.0765
Share to
QR

Scan QR to access full text

Cite this article
BibTeX
Citations
表12种不同金属材料的力学参数
科
Family		 属数
Number of
genus		 种数
Number of
species		 占总种数比例
Percentage of
total species (%)		 属
Genus		 种数
Number of
species		 占总种数比例
Percentage of total
species (%)
鹅膏菌科Amanitaceae 2 11 5.26 鹅膏菌属 Amanita 10 4.78
小菇科 Mycenaceae 2 12 5.74 丝盖伞属 Inocybe 5 2.39
多孔菌科 Polyporaceae 8 14 6.70 蜡蘑属 Laccaria 5 2.39
红菇科 Russulaceae 3 23 11.00 小皮伞属 Marasmius 6 2.87
小菇属 Mycena 11 5.26
光柄菇属 Pluteus 5 2.39
红菇属 Russula 17 8.13
栓菌属 Trametes 5 2.39
关闭全屏
  • BibTeX
  • EndNote
  • RefWorks
  • TxT
Links
Articles
Latest Articles
Most Read
Collections
Updates
Events
News
Multimedia
About
About Us
Contact
No. 86 Xueyuan South Road, Haidian District, Beijing
100081
010-62662699(Operation and Technical Support)
qkjq@cast.org.cn

Copyright © 2025 China Association for Science and Technology. All rights reserved. For all open access content, the relevant licensing terms apply.
Sponsored by the Office of the Leading Group for Cybersecurity and Informatization of CAST, and supported by Science and Technology Review Publishing House