Researchers are calling for the application of community-led strategies, supported by stronger local evidence, to improve early detection and targeted treatment of cancer. The focus is on how diagnostics, biomarkers, and artificial intelligence can be tailored to the local needs of specific populations.
At a symposium titled “Cancer Research: Genomics, AI, and Targeted Therapies” Experts explored the full spectrum of cancer research, from prevention and screening to routine laboratory tests, biomarker discovery, and AI-assisted diagnostics. Discussion centered on the need to adapt these advances to the unique symptoms and clinical needs of local patient populations.
The symposium, which included researchers from a variety of countries, focused on cancer patients in the United Arab Emirates (UAE), but its broader message was clear. The idea is that improved outcomes are achievable if screening, detection, and treatment strategies are informed by local data-based evidence and research.
The two-day symposium (May 20-21) brought together cancer researchers, clinicians, laboratory experts, computer scientists, and drug discovery experts from the UAE, Spain, and Russia. Discussion focused on breast, thyroid, colorectal, and prostate cancer, reinforcing the important argument that screening programs must be linked to local research. Instead, local studies should help explain why cancer patterns in the UAE differ from those observed in other regions.
Professor Riyad Bendardaf, medical oncologist and director of the Cancer Research Center at the University of Sharjah, stressed that cancer patterns in the UAE do not necessarily reflect global trends, and stressed the need to adopt specific prevention strategies that meet the needs of the local population.
“If we had the right screening tools and the right diagnostic tools, the burden in the UAE would be reduced by 40%,” Professor Bendarduff said in a presentation on the global impact of cancer. “Through effective screening programs for these three cancers, (we) will reduce the burden of this disease by 56.7%,” he added, referring to breast, thyroid and colorectal cancers.
He noted that the UAE’s cancer profile follows global patterns, with thyroid cancer particularly prominent in local statistics. “This is not seen globally, so why is the thyroid second? Our numbers are different from the international numbers,” he said.
According to Professor Bendalduff, these differences should directly inform how screening strategies should be designed. Highlighting local data on female cancer patients, he pointed out that “43.8% of female cancer patients in our society are under the age of 50.”
This distribution of young people requires us to reconsider when to start screening, he explained. “So we need to start testing 10 years earlier than 40 years old,” he said. “We are getting cancer 10 years earlier than Western societies. We need to adjust based on the data and target screening at younger ages.”
Laboratory tests and biomarkers
Speakers also emphasized the growing importance of routine clinical testing, biomarker discovery, and molecular diagnostics and their central role in monitoring cancer and guiding treatment decisions.
Dr. Noura Ali Alikhayal, Head of Laboratory at Sharjah University Hospital, UAE, outlined the role of routine and advanced laboratory tests throughout cancer treatment, from initial assessment and treatment monitoring to survival and follow-up. “Regular laboratory tests remain an important cornerstone of cancer monitoring and treatment safety,” she stressed.
She explained how standard blood and biochemical markers can help clinicians assess disease burden, detect treatment toxicity, and identify complications early. For example, hemoglobin levels can help make decisions about disease severity, toxicity and tolerance of disease treatments, and the need for blood transfusions. Lymphocyte counts can assess immune status, and elevated neutrophil-to-lymphocyte ratios are associated with poor prognosis.
Dr. Alikhayal further emphasized the importance of electrolytes and metabolism in tumor treatment, describing electrolyte imbalance as “one of the most common metabolic complications in oncology patients.” Such disruptions could be due to tumor biology, treatment effects, or systemic disease processes, she noted.
Aggressive electrolyte monitoring and early intervention are essential to prevent life-threatening complications during cancer treatment. ”
Dr. Noura Ali Alikhayal, Director, Sharjah University Hospital Research Institute
The search for earlier and more accurate cancer markers was also emphasized in the discussion on colorectal cancer. Professor Rifat Hammoudi, Director of the Institute of Medical and Health Sciences at UoS, presented research on early-stage biomarker identification using convolutional neural networks, an advanced form of artificial intelligence designed to analyze images and patterns.
“Colorectal cancer is one of the top three cancers in the world,” he said. “The incidence is high and the mortality rate is high,” he said, stressing that early diagnosis of the disease is a major challenge, as the symptoms are often difficult to detect in the early stages. “Early detection is not easy because of the symptoms,” he explained, emphasizing the importance of developing reliable early biomarkers to expand treatment opportunities.
Professor Hammoudi’s presentation also reflected the broader theme of the symposium: the need for researchers from different fields to collaborate in cancer research. “What’s unique about UoS is that we have people from pharmacy, health sciences, oncology and pathology backgrounds,” he said. “So if we work together, we may find more meaningful answers.”
He warned that narrowing the focus of the field would limit understanding. “If we only focus on our field, we may not actually fully understand cancer because it’s so complex,” he added.
Use of AI in oncology and clinical practice
Artificial intelligence emerged as a key theme at the symposium, but speakers repeatedly cautioned against treating it as a simple solution. Instead, they emphasized the need for rigorous validation, interpretation, and clinical relevance.
Professor Hammoudi pointed to the inherent complexity of cancer patterns and highlighted the challenges of applying AI to histopathology and molecular data. Histopathology, the microscopic examination of cancer tissue, is essential for cancer diagnosis and staging, but it remains difficult for AI systems to provide reliable interpretation.
“The reason AI doesn’t work is because computers aren’t intelligent enough to actually recognize highly complex, highly intertwined patterns,” he explained.
At the same time, we demonstrated how a more carefully designed AI approach can improve performance. He discussed the use of explainable AI, a technique that allows researchers to understand how AI models draw conclusions. He also demonstrated how Grad-CAM, an explainable AI technology, can aid in decision-making. According to his presentation, this approach achieved 99.1% classification accuracy in distinguishing between ulcerative colitis, colorectal cancer, and normal colon tissue.
The role of AI in digital pathology was examined from a clinical perspective. Dr. Alikhayal said AI-powered pathology can help with automated tumor detection, biomarker prediction, outcome prediction, and faster workflows. However, she noted that digital pathology and AI are now more common in research settings than in routine clinical practice, but this is largely due to validation limitations.
The gap between promise and clinical readiness has been a recurring theme, especially in discussions around generative AI. Dr. Svetlana Illarionova, Head of Skoltech’s Computer Vision Research Group, presented her research on virtual staining in histopathology using generative AI.
Her research explored the use of AI to generate immunohistochemistry-like images from standard H&E-stained slides. While H&E staining reveals the overall tissue structure visible under a microscope, immunohistochemistry uses antibodies to detect specific proteins within the tissue. For example, in breast cancer, HER2 staining can help identify whether cancer cells have high levels of the protein, which can influence treatment choices.
Dr. Ilarionova compared different AI approaches, including GAN-based methods and diffusion models. Diffusion models can produce higher quality images, but also require more computational resources.
Importantly, she emphasized that clinical use requires more than images that appear visually convincing. “Looking right and being right are not the same thing,” she warned.
AI-generated virtual staining can distort the size of cells, blur or fuse structures, shift cell positions, and create features that don’t exist, she explained. For pathologists who rely on precise cellular details to make diagnostic decisions, such inaccuracies pose a significant barrier to effective clinical decisions.
For virtual staining to be clinically meaningful, she argued, future systems will need to incorporate pathology expertise, preserve cell structure, use expert evaluation, and be trained on more powerful datasets. “General image-to-image translation is insufficient for clinical work,” she said.
Dr. Maxim Sharaev, assistant professor and director of applied AI research at Skoltech, expanded the discussion with a presentation on the role of generative AI in precision oncology. While highlighting the rapid adoption of AI-enabled medical devices, particularly in the field of medical imaging, he cautioned against viewing generative AI as a direct clinical decision-maker.
“Illusions are inevitable,” he said, warning that generative AI, while plausible, can produce output that is inaccurate, misleading, and not based on real data. For this reason, such systems should not be used directly for clinical diagnosis, he said.
Rather, Dr. Sharaev argued that the most valuable and responsible application of generative AI is in supporting research. This includes integrating complex and multimodal forms of patient data, generating synthetic datasets for model training, and embedding expertise into AI systems.
He presented a research collaboration focused on the use of generative AI to integrate different types of patient data, including pathological images and molecular profiles. RNA sequencing (RNA-seq) reveals gene activity within a sample, and DNA methylation data captures chemical modifications that regulate gene expression. By combining these data sources, researchers can obtain a more detailed picture of a patient’s cancer.
Another research stream is exploring the use of generative models to create synthetic pathological image data from healthy MRI scans. This approach addresses the lack of labeled disease data and integrates model training.
“The images generated will not be used directly in the clinic,” he stressed. Rather, they can be used as training models to improve performance, especially in difficult detection tasks.
