A research revolution: multiomic and spatial techniques in cancer biology

When we visited AACR 2025 (24-30 April 2025; Chicago, IL ,USA) it rapidly became clear that two developing technological fields were taking the conference by storm: multiomic and spatial techniques. The spatial resolution of biological data is an exciting emerging area in cancer research, which involves studying the organization of tumor architecture, cell types, and protein expression within a tissue sample. By understanding the spatial arrangement, researchers can better grasp why a drug might work, or fail, in a specific patient. Meanwhile, multiomic techniques enable researchers to make near instant connections between fluctuations at different levels of the central dogma of molecular biology and their role in numerous aspects of cancer. Combining these spatial and multiomicmethods with AI and machine learning could allow for a deeper understanding of tumor growth and drug mechanisms. In this article, we reveal insights from the many researchers we caught up with at the event on how they use mulitomic and spatial techniques to understand the tumor microenvironment, predict drug responses and personalize future care and also share their tips and tricks for working with these technologies. Multimodal foundation models in personalized cancer care Benjamin Haibe-Kains, Princess Margaret Cancer Centre (ON, Canada), highlights that multiomic and spatial profiling technologies are transforming cancer research by offering an unprecedented view of tumor biology. While these tools are currently costly and complex, they are able to provide insights that could train predictive models based on more accessible clinical data, such as pathology images and blood tests. This process of "knowledge distillation" helps bridge the gap between cutting–edge discovery and practical clinical application, moving research findings into personalized cancer care more quickly. Haibe-Kains also believes that multimodal foundation models, highly flexible AI models trained on vast amounts of diverse data, could revolutionise personalized medicine. He shares that they could one day be utilized to predict therapy resistance, uncover new biomarkers, and classify patients based on their unique molecular and spatial features. This ability to generalize across different data types and clinical contexts could significantly accelerate both fundamental and translational research. To make this evolution possible, Haibe-Kains advises organizing data with future AI and machine learning applications in mind, ensuring not only high–quality data but also rich contextual information like biopsy location and tumor subtype. He highlights the challenge of ensuring that findings from these technologies are reproducible and generalizable across different biopsies and tumor regions, and emphasizes the clear need to design studies that prioritize both deep biological understanding and real–world applicability. AI-based fragmentomic approach could turn the tide for ovarian cancer Learn more about a new combined fragmentomic–protein biomarkers test could change the prognosis for ovarian cancer. Integrating AI and multiomics into computational cancer biology Maria Secrier, University College London (UK), leads a team developing machine learning and data integration methods focused on understanding how cancer cells evolve. The recent "spatial revolution" is enabling unprecedented tissue profiling, but she believes the field is still in its early stages of truly understanding how spatial cellular behavior contributes to cancer development. There's a growing interest in defining functional spatial niches, and an exciting trend is the use of AI to integrate data across multiple scales, including genomics, transcriptomics, and imaging. While these approaches can predict clinical outcomes, new methods are needed to improve their interpretability. Secrier provides several key tips for researchers, starting with the critical importance of quality control and careful annotation of both cell types and their dynamic states. She also advises close collaboration between computational and experimental labs from the beginning, ensuring that computational findings can inform experimental design and validation. When using public data, researchers should be realistic about its limitations. She also stresses the need to make all analyses fully reproducible and to share data to benefit the wider community. Finally, she suggests that researchers should not be too quick to disregard outliers, as they may hold the key to unexpected discoveries. Integrative omics for characterizing addiction and brain development In this interview, Melyssa Minto, RTI International (NC, USA), discusses her research, the challenges of integrating complex, multi-layered data, the future of multiomics and what it means to be a multiomic bioinformatician. Spatial techniques enable computational analysis of tumor heterogeneity For Itay Tirosh, Weizmann Institute of Science (Israel), three techniques are better than one. His team is unearthing new insights into tumor heterogeneity by combining computational methods with single–cell and spatial technologies. Tirosh believes that although the field of spatial transcriptomics is a rapidly growing area, with large amounts of data being generated from many competing technologies, real insights are starting to emerge. Tirosh highlights a recent session he was involved with at AACR 2025, discussing the tumor microenvironment and how spatial techniques were leading to new discoveries in this area. This discussion highlighted two key findings. First, all speakers acknowledged the existence of consistent TME patterns where specific cell types tend to be colocalized across numerous tumors. While unclear what generates these patterns and their implications, it has been revealed that colocalization patterns demonstrate stronger associations with clinical features than the mere abundance of specific cell types and states, making the spatial data vital for clinical assessment. Second, hypoxia has been identified as a driver of increased spatial structure. For instance, in glioma specifically, tumor regions near hypoxic areas show greater organization such that cells tend to be surrounded by cells similar to them, while five distinct tumor layers, reflecting distance from hypoxia, are formed. When applying spatial techniques to studying cancer, he shares that his key tip is to avoid overgeneralizing observations from individual tumors, as each one has a unique spatial pattern. Instead, the goal should be to identify common patterns across many tumors. Understanding these patterns and what generates them is an ongoing challenge, but initial findings show that these colocalization patterns are more strongly associated with clinical features than the simple abundance of specific cell types. Tirosh highlights that advanced computational methods are essential for extracting meaningful insights from the sheer size and complexity of spatial data and picks out the Visium HD platform as a particularly impressive technology, due to its data quality and resolution, alongside its ability work with archival FFPE samples. Computational approaches for spatial studies Spatial techniques have continued to rise in popularity in recent years; but what computational approaches are available for spatial studies? This article, by no means an exhaustive list, details a few. Spatial technology and radioligand therapy Shiva Malek is the Global Head of Oncology Disease Area at Novartis BioMedical Research (MA, USA), where her focus on imaging and spatial techniques is linked to an interest in radioligand therapies. These therapies are often paired with imaging approaches, such as SPECT or PET, to visualize where the drug goes within the body. This allows scientists to see the drug's distribution in both normal and tumor tissues and to calculate the absorbed radiation. To those wishing to apply these techniques themselves, Malek advises first defining the specific research question and then selecting the right model for studying target distribution. She highlighted that it’s crucial to have a clear understanding of the basic principles, such as whether a particular antibody or peptide will cross–react with different species, to ensure you interpret your imaging data accurately. Malek highlights that combining spatial technology with imaging can be “really powerful; there is no other technology platform or modality where you can both see where the drug is going and understand its pharmacodynamic effect”. The signals underlying the spatial: key techniques evolving cancer research We spoke to Shiva at the annual meeting of the American Association for Cancer Research in Chicago to learn more about her award, the techniques she uses to investigate cancer cell signaling and the tumor microenvironment, and the methods she’s most excited to see implemented in cancer research. Uncovering the secrets of the intratumoral microbiome Alice Martin's (Francis Crick Institute; London, UK) poster on kidney cancer progression earned her the AACR 2025 Women in Cancer Research Scholar Award. Her work focuses on the intratumoral microbiome as a potential new target for cancer treatment. To gather her data, she leveraged whole–genome sequencing and bulk RNA-seq to extract bacterial reads from tumor samples, and stressed the importance of “careful denoising and decontamination” to avoid misinterpreting human DNA as bacterial. After bacterial colonies of Cutibacterium were cultured from the tumor samples to confirm the presence of live bacteria, Martin utilized RNAscope on patient tumors to visually confirm the presence of Cutibacterium within the tissue. To keep the data clean, Martin recommends a human reference genome at every step and sequencing water samples that have gone through the same process to remove common lab contaminants. She also suggests confirming bacterial presence through both sequencing and culturing, and optimising your signal–to–noise ratio when visualizing sparse signals by using fresh frozen tissue for RNAscope. Women in Cancer Research Scholar Award: The intratumoral microbiome as a novel therapeutic target Alice Martin discusses her Women in Cancer Research Scholar Award for her work on kidney cancer progression. Replicating tumors with bioprinting A new method for creating patient–specific tumor replicas using 3D bioprinting could allow researchers to control the exact location and type of cells, a significant advance over traditional organoid models. Haylie Helms, Oregon Health and Science University (OR, USA) is developing just such a method, which begins with a patient biopsy. This is then analyzed to create a detailed map of the cell types present and their spatial organization within different "neighborhoods" of the tumor. Using this map, a 3D printer deposits different cell types, such as cancer and immune cells, into a "tumor avatar" that is physiologically comparable to the original biopsy. Helms believes her method could overcome the limitations of conventional organoids and spheroids, where researchers can only control the number and ratio of cells, relying on them to self–organize. With this new bioprinting platform, the precise geometry and spatial relationships of the cells can be reproduced. This is critical for studying cell–to–cell interactions and how a drug's effectiveness might be influenced by a tumor's architecture. She hopes this technology will be a valuable tool for a wide range of cancer research, complementing existing methods to study tumor evolution and test drug efficacy in a controlled and reproducible environment. Bioprinting 3D spatially resolved tumor avatars to mimic the native tumor microenvironment In our pick of the AACR posters, Haylie Helms discusses her poster and how this work could lead to real-world benefits. Unlocking the tumor microenvironment with single–cell spatial data A new method In a new investigative approach, Tianyou Luo (Tempus AI, IL, USA), leverages a neural network to process spatial transcriptomics data from the 10x Genomics Visium HD platform, mapping it to single–cell resolution. This novel method addresses a key limitation of the Visium HD platform's default 8 µm resolution, which can merge multiple cells or split single cells into different data bins. By achieving single–cell resolution, this method could allow researchers to apply standard single–cell RNA–seq analysis techniques and gain a deeper, more accurate understanding of the TME. Utilizing his method, Lou has achieved successful cell-level binning of gene counts in non-small cell lung cancer samples. The resulting cell clusters, annotated with the help of a large language model, have shown high agreement with a pathologist's annotations, correctly identifying key cell types. He hopes this technique will improve biomarker discovery and provide new insights for immuno–oncology and antibody–drug conjugate research. He highlights that his method requires high–quality H&E images for successful segmentation and cautions the data's inherent sparsity should be considered. Visium HD combined with deep-learning-based cell segmentation on H&E images yield accurate cell annotation at single-cell resolution Tianyou Luo discusses his AACR poster and he hopes it will advance current knowledge about the TME. Unveiling therapeutic targets and genetic drivers forming cancer's ecosystem Di Zhao, The University of Texas MD Anderson Cancer Center (TX, USA), is a cancer biologist using multiomic and spatial techniques to characterize intrinsic genetic changes in cancer cells that enable them to influence their microenvironment and develop combination treatments for personalized medicine. Zhao’s research demonstrates how these methods could also help identify molecular features that can predict a patient's response to therapy. Her team has successfully identified therapeutic targets in different subtypes of prostate cancer, providing strong evidence of their potential in preclinical models. These technological advances are fundamentally changing our understanding of cancer and its interactions with the body's ecosystem at every stage. She highlights that single–cell and spatial transcriptomics generate vast amounts of information for assessing correlations and forming hypotheses, but it is critical to integrate different multiomic approaches – such as genetic, transcriptomic, and proteomic analyses – to identify the most important findings. She also stresses the need for functional validation of these observations in appropriate in vitro and in vivo systems to distinguish between what is a cause and what is a consequence of cancer progression. How is Spatial Transcriptomics Influencing Cancer Research and Diagnostics? Spatial transcriptomics has continued to rise in popularity since it was developed in 2016; but how is it being utilized in tumor biology and diagnostics research? Computational advances optimizing flow matching A key trend in cancer research is the integration of high-resolution multiomics and spatial data, particularly through the use of computational methods. One promising development is Jovan Tanevski’s (Heidelberg University Hospital, Germany) work on the use of optimal transport and flow matching to integrate different types of data, such as transcriptomics and proteomics. An example of this is DOT, which combines information from both reference data and spatial omics data to accurately align and integrate datasets. Tanevski is also developing another computational tool: topography-aware oriented spatial omics analysis tool (TOAST). TOAST can match tissue slices from the same sample or different samples, and it works with various omics technologies. By aligning data in this way, researchers can transfer information between slides and collect information from different samples. Tanevski has demonstrated that TOAST can be utilized for both intra– and inter–sample alignment in transcriptomics and proteomics, and has also been shown to align temporal data, helping to track how cells move or change states over time. Tanevski highlights that he and his team have designed these open-source tools to be user–friendly, with specific instructions and examples. He believes that the true value of these methods lies in their ability to answer biological questions that cannot be solved with other technologies alone. From pixels to patients: tools for the integration of spatial omics data At AACR 2025, we caught up with Jovan to discuss his session at the meeting, the tools he’s developed for integrating spatial omics data and what the computational community should focus on in cancer research.