Unveiling breast cancer heterogeneity by integrating computational approaches and multi-omics data

Vollständiger Abstract

Worum geht es in dieser Arbeit?

Breast cancer is a heterogeneous disease, exhibiting significant variations both between patients and within individual tumors. This complexity, driven by diverse genetic and transcriptomic alterations, poses a significant challenge for accurate diagnosis and effective treatment in clinic settings.In paper I, we benchmark intrinsic molecular subtyping by developing BreastSubtypeR, a unified computational tool for the research field that integrates ten established molecular classification methods. Its innovative "AUTO" mode dynamically selects the most appropriate methods for a given dataset, minimizing user bias and enhancing the accuracy, consistency, and reproducibility of intrinsic subtyping.In paper II, we enhance the resolutions of spot-based spatial transcriptomics data (e.g., 10x Visium) by developing a computational tissue annotation (CTA) pipeline. This method leverages machine learning to perform high-resolution, automated annotation of histology images from fresh-frozen tissues, precisely mapping tumor and stromal compartments onto the spatial transcriptomics data. This integration provides crucial information on cellular context for gene expression analysis.In paper III, we leverage these tools to dissect intra-tumoral heterogeneity in progressed, treatment-naïve large breast tumors. Through the integration of data from bulk whole-genome/exome sequencing, methylation, and RNA sequencing, including high-resolution spatial transcriptomics (10x Visium and Xenium) in a multi-region setting, we uncovered a complex tumor ecosystem with distinct spatial distributions, also known as niches. These niches are driven by underlying subclonal populations and are characterized by a unique tumor microenvironment. The high-resolution spatial transcriptomic data offer a detailed view of dynamic tumor evolution and ecosystem diversity.In summary, this thesis progresses from providing essential computational resources for standardized classification and high-resolution spatial annotation to deploying them in a comprehensive multi-omics investigation. Our work reveals the intricate and dynamic heterogeneity of breast cancer and underscores that appreciating this internal diversity is paramount for advancing personalized treatment and overcoming therapeutic resistance.List of scientific papersI. Qiao Yang, Johan Hartman, Emmanouil G Sifakis. BreastSubtypeR: a unified R/Bioconductor package for intrinsic molecular subtyping in breast cancer research. NAR Genomics and Bioinformatics. 7, 4 (2025). https://doi.org/10.1093/nargab/lqaf131II. Tianyi Li, Qiao Yang, Balazs Acs, Emmanouil G. Sifakis, Hosein Toosi, Camilla Engblom, Kim Thrane, Qirong Lin, Jeff E. Mold, Wenwen Sun, Ceren Boyaci, Sanna Steen, Jonas Frisén, Jens Lagergren, Joakim Lundeberg, Xinsong Chen ** , and Johan Hartman **. Computational pathology annotation enhances the resolution and interpretation of breast cancer spatial transcriptomics data. NPJ Precis. Onc. 9, 310 (2025). https://doi.org/10.1038/s41698-025-01104-3III. Qiao Yang, Tianyi Li*, Emmanouil G. Sifakis*, Ran Ma, Jeff E. Mold, Qirong Lin, Hosein Toosi, Kim Thrane, Shi Yong Neo, Ira Oikonomou, Irma Fredriksson, Wenwen Sun, Yuqing Zhou, Sonia Corral Leal, Rapolas Spalinskas, Katarina Tiklova, Joakim Lundeberg, Jens Lagergren, Jonas Frisén, Camilla Engblom, Xinsong Chen ** , and Johan Hartman **. Intratumoral heterogeneity of treatment-naïve large breast tumors unveiled by multi-region sequencing and multi-omics characterization. [Manuscript]* These authors contributed equally** These authors contributed equally