The starting point for the study is a familiar frustration in respiratory medicine. Asthma is a heterogeneous condition characterized by airway inflammation and obstruction driven by edema, bronchospasm, and excessive mucus production, but the genetic factors responsible for its pathogenesis remain largely unknown despite decades of progress. Reliable biomarkers are essential for diagnosis, for understanding disease mechanisms, and for predicting how individual patients will respond to treatment. Existing biomarkers capture only part of this picture, and the authors argue that new ones are needed to refine treatment strategies and guide therapeutic decisions over the long term.

To find them, the team turned to the Gene Expression Omnibus (GEO), the public repository maintained by the US National Center for Biotechnology Information. Three datasets — GSE64913, GSE67472, and GSE137268 — supplied the raw material. These datasets contain gene expression profiles sampled from the airway epithelium and from induced sputum of people diagnosed with asthma, providing complementary views of the tissue environments where the disease manifests. By analyzing samples drawn from these two sources, the researchers could look for genes whose activity correlates strongly with asthma status across different sampling contexts, a strategy designed to filter out artifacts tied to any single tissue type or collection protocol.

The analytical pipeline itself unfolds in three stages, each addressing a distinct computational challenge. The first stage uses a pretrained convolutional neural network (CNN) for feature extraction. CNNs, best known for their success in image recognition, apply layers of learned filters that abstract increasingly abstract representations from raw input. Here, the network converts high-dimensional gene expression profiles into new deep features — compact numerical descriptions that capture patterns of co-regulation and differential expression that simpler statistical tests might miss. Transfer learning from a pretrained model means the network arrives already equipped with general-purpose pattern detectors, which the authors then adapt to the specific structure of transcriptomic data.

Deep feature extraction, however, creates its own problem: dimensionality. Gene expression datasets typically measure tens of thousands of genes across comparatively few samples, and adding CNN-derived features can make the search space even more unwieldy. The second stage of the pipeline confronts this with a modified black winged kite (MBWK) algorithm, an enhanced version of a metaheuristic optimization method inspired by the hunting behavior of the black-winged kite, a bird of prey that hovers and strikes with precision. Metaheuristics of this kind explore large solution spaces by iteratively adjusting a population of candidate solutions, balancing exploration of new regions against exploitation of promising ones. In this application, the algorithm selects the optimal subset of features, discarding redundant or noisy dimensions while preserving the signals most informative for distinguishing asthmatic from healthy expression profiles.

The third and most novel stage applies an adaptive dynamic graph neural network (AD-GNN) to the optimized features. Graph neural networks represent data as networks of nodes and edges, and they are naturally suited to genomics because genes do not act in isolation — they operate within regulatory networks, signaling pathways, and protein interaction maps. An AD-GNN can adapt its structure dynamically as it learns, propagating information across the gene interaction graph so that the activity of any single gene is interpreted in light of its neighbors. This relational reasoning is what allows the model to identify hub genes: highly connected genes whose expression patterns sit at the center of asthma-associated network activity, rather than merely differing on average between patient groups.

The output of this three-stage process is a short list of candidate genes: CCDC167, SERPINB2, SEC14L1, and POSTN. Each has a plausible biological story. POSTN, which encodes the matricellular protein periostin, has long been associated with type 2 airway inflammation and epithelial remodeling in asthma, lending the pipeline immediate face validity. SERPINB2 belongs to the serpin family of protease inhibitors and is known to respond to inflammatory signaling in epithelial tissue. CCDC167, a coiled-coil domain-containing protein of less certain function, has recently attracted attention as a potential biomarker for airway inflammation, and its independent emergence here reinforces that hypothesis. SEC14L1, involved in lipid handling, rounds out the set. The authors report that the predicted genes demonstrate functional relevance across a variety of asthma-related tissues, supporting their candidacy as genuine disease markers rather than statistical coincidences.

What makes the study notable is less any single component than the way the components interlock. CNN feature extraction addresses the problem that raw expression values are noisy and high-dimensional; the MBWK optimizer addresses the problem that even informative features come bundled with redundancy; and the AD-GNN addresses the problem that genes must be understood in context. The authors report that the proposed model successfully identified both asthma biomarkers and treatment options as demonstrated by their experiments, positioning the framework as a template for biomarker discovery that could in principle be retrained on other diseases by swapping in different datasets. Because all three GEO datasets are publicly available, the entire approach is reproducible by any laboratory with the computational resources to run it.

The broader context is a rapidly growing effort to bring artificial intelligence into precision medicine for respiratory disease. Recent years have seen machine learning applied to cough sound analysis for automated asthma detection, to routine blood biomarkers for asthma prediction, to the prediction of allergen immunotherapy response, and to the identification of ferroptosis-related and cellular senescence-related genes in severe asthma. Omics approaches — genomics, transcriptomics, proteomics, metabolomics — generate data at a scale that outstrips classical statistical methods, and deep learning offers a way to extract signal from that abundance. The Kakatiya University study sits squarely within this movement, and its use of graph-based learning reflects a growing recognition that biological data are inherently relational.

Cautions remain, as they always do at this stage of translational research. Candidate biomarkers identified computationally must be validated experimentally — at the protein level, in independent patient cohorts, and ultimately in prospective clinical studies — before they can inform diagnosis or drug development. The authors themselves used publicly available, de-identified data with no direct human participation, which means the pipeline’s outputs are hypotheses to be tested rather than clinical conclusions. Still, the study offers a concrete demonstration that a carefully engineered chain of deep learning tools can compress a genome-wide search into a handful of testable targets. For a disease as heterogeneous and burdensome as asthma, tools that accelerate that search from years to months could meaningfully change how quickly new diagnostics and therapies reach patients, and the four genes highlighted here now join the growing list of leads awaiting experimental confirmation.

Subject of Research: Deep learning-based gene expression analysis for identifying diagnostic biomarkers and therapeutic targets in asthma

Article Title: Deep learning-based analysis of gene expression for identifying diagnostic biomarkers and therapeutic targets in asthma

Article References: Iqbal, M. A., Sadanandam, M., & Kumaraswamy, A. (2026). Deep learning-based analysis of gene expression for identifying diagnostic biomarkers and therapeutic targets in asthma. Neural Computing and Applications, 38(19), Article 768. https://doi.org/10.1007/s00521-026-12519-9

Image Credits: AI Generated

DOI: 10.1007/s00521-026-12519-9

Keywords: asthma, deep learning, gene expression, biomarkers, graph neural networks, convolutional neural networks, GEO database, therapeutic targets, precision medicine, bioinformatics, airway epithelium, machine learning

Cardiovascular diseases remain the leading cause of mortality globally, despite decades of advancements in clinical management and pharmacology. Existing predictive models primarily rely on clinical risk factors such as blood pressure, cholesterol levels, age, and lifestyle indicators but often lack the granularity to account for complex biological heterogeneity between patients. The advent of high-throughput omics technologies—genomics, transcriptomics, proteomics, metabolomics, and epigenomics—offers a treasure trove of molecular data that can capture disease mechanisms at multiple biological layers. Integrating these data streams, however, poses significant analytical challenges due to their high dimensionality, heterogeneity, and the nonlinear interactions inherent in biological systems.

The study harnesses state-of-the-art artificial intelligence methodologies, including deep learning architectures and advanced feature integration algorithms, to fuse multiomics signals from large patient cohorts. By doing so, the model identifies subtle, nonlinear patterns that escape traditional statistical techniques. Notably, the AI framework does not treat each omics layer in isolation but treats them complementary—each providing unique and overlapping information that together creates a holistic molecular portrait of cardiovascular risk. This integrative approach surpasses the predictive power of any single omics dataset or conventional clinical models by a significant margin.

Luo and colleagues first assembled an extensive multiomics dataset comprising whole-genome sequencing, RNA expression profiles, circulating proteome, metabolite panels, and epigenetic modifications from thousands of individuals with varying cardiovascular outcomes. Such rich data allowed them to interrogate the pathophysiology of CVD at unprecedented depth. The AI model was then trained and validated using classical cross-validation alongside external cohort testing to ensure robustness and generalizability. The multiomics-enabled AI consistently delivered superior accuracy in predicting adverse cardiovascular events compared to established clinical calculators like the Framingham Risk Score or ASCVD risk estimator.

One of the key innovations in this study is the use of interpretable AI techniques to elucidate which omics features most critically contribute to risk prediction. Genetic variants associated with lipid metabolism, gene expression signatures indicative of inflammatory pathways, proteomic markers related to vascular remodeling, and specific metabolite fingerprints emerged as dominant contributors. This layered insight not only enhances predictive accuracy but also unravels potential mechanistic underpinnings that may be targeted for therapeutic interventions. The study bridges the gap between ‘black-box’ AI predictions and biologically meaningful interpretations, a crucial step towards clinical adoption.

Moreover, the researchers demonstrated that integrating omics layers provided synergistic benefits. For example, certain genomic risk loci were only predictive in the context of specific transcriptomic profiles, highlighting gene-environment and gene-gene interactions captured through molecular phenotypes. Metabolomic data further refined risk stratification by reflecting real-time biochemical alterations, while epigenomic markers offered clues about gene regulation dynamics affected by lifestyle and environmental exposures. Such multi-dimensional profiling advances our understanding from static snapshots to dynamic molecular ecosystems relevant to disease progression.

Importantly, the AI-driven multiomics model excels in identifying at-risk individuals who might be missed by traditional screening methods. This has profound implications for early diagnosis and intervention where timely lifestyle changes or preventive therapies can radically alter disease trajectories. Personalized risk assessments can be dynamically updated as new omics data becomes available, allowing continuous refinement of prognostic accuracy. The study underscores the feasibility of implementing such systems in clinical workflows, leveraging advances in high-throughput molecular assays and computational infrastructure.

The translational potential extends into the realm of drug development and precision therapeutics. By highlighting distinct molecular signatures linked to subtypes of cardiovascular disease, the AI model paves the way for stratified clinical trials and targeted treatments. Biomarkers discovered through this integrative approach might serve as companion diagnostics or surrogate endpoints, accelerating regulatory approval processes. Furthermore, understanding the molecular basis of cardiovascular risk at multiple omics levels may uncover novel therapeutic targets inaccessible through single-layer studies.

Despite these promising breakthroughs, the authors emphasize challenges and future directions. Standardizing multiomics data acquisition, harmonizing batch effects, and ensuring longitudinal data availability are critical for clinical utility. Privacy concerns surrounding comprehensive molecular profiling necessitate secure data-sharing frameworks and ethical guidelines. Additionally, expanding cohort diversity is imperative to prevent algorithmic biases and ensure equitable healthcare benefits across populations. Ongoing improvements in AI interpretability, computational efficiency, and integration with electronic health records will further catalyze real-world adoption.

This study by Luo et al. marks a paradigm shift in cardiovascular risk prediction by demonstrating the power of AI-based multiomics integration. The authors’ visionary approach offers a comprehensive molecular lens through which the complexity of cardiovascular disease can be unraveled and addressed on an individual basis. As biomedical technologies continue to evolve, such interdisciplinary synergy between AI and omics sciences holds the promise to transform our approach to one of humanity’s most pressing health challenges, undoubtably steering us closer to the long-sought goal of truly personalized medicine.

In summary, the integration of multiomics datasets with advanced AI analytics establishes a robust predictive framework that transcends the limitations of traditional clinical models. By revealing complementary contributions from genomics, transcriptomics, proteomics, metabolomics, and epigenomics, this approach creates a nuanced and dynamic map of cardiovascular risk factors. The deep biological insights emerging from this work enrich our understanding of disease etiology, while offering actionable intelligence for prevention, diagnosis, and therapeutic interventions. As these technologies mature and become increasingly accessible, they promise to revolutionize cardiovascular healthcare on a global scale.

Looking ahead, collaborative efforts to expand multiomics databases, refine AI algorithms, and experimentally validate molecular findings will be critical. Integrating real-world clinical data with molecular profiles promises continual model refinement, driving precision medicine into routine practice. This transformative research underlines how the fusion of AI and multiomics heralds a new era in biomedicine—one where the complexity of human biology is decoded to deliver personalized, predictive, and preventive healthcare tailored to each individual’s unique molecular blueprint.

Subject of Research: AI-based multiomics profiling for personalized prediction of cardiovascular disease.

Article Title: AI-based multiomics profiling reveals complementary omics contributions to personalized prediction of cardiovascular disease.

Article References:
Luo, Y., Zhang, N., Yang, J. et al. AI-based multiomics profiling reveals complementary omics contributions to personalized prediction of cardiovascular disease. Nat Commun (2026). https://doi.org/10.1038/s41467-026-68956-6

Image Credits: AI Generated

Colorectal cancer remains one of the most pressing health challenges globally, necessitating advancements in diagnostic techniques that can evolve alongside our understanding of cancer biology. The Tumor-Stroma Ratio is a critical parameter in cancer prognosis, as it correlates significantly with patient outcomes. In the context of colorectal cancer, accurately distinguishing between tumor and stroma regions can delineate between aggressive and indolent disease forms. This integrative approach using machine learning aims to refine the precision of these assessments, contributing greatly to personalized patient management strategies.

The integration of deep learning into the analysis of histopathological slides represents a paradigm shift from conventional methods. Traditional manual assessments are not only labor-intensive but also subject to variances stemming from pathologist experience and subjective interpretation. By applying deep learning techniques that use patch-based classification and segmentation, this research seeks to mitigate these issues. The proposed Efficient-TransUNet model adeptly classifies patches of tissue as either normal or abnormal while concurrently segmenting critical tumor and stroma regions.

As the research reveals, the outcomes achieved through this advanced methodology significantly exceed those obtained from traditional assessment techniques. The model’s ability to automate the TSR computation is not merely a technological triumph; it represents an essential leap towards improving diagnostic workflows. The enhanced objectivity and consistency provided by the automated approach support increased diagnostic reliability, which is crucial in clinical settings where timely decisions must be made.

One of the standout features of the Efficient-TransUNet is its ability to effectively differentiate between stroma-high and stroma-low tumors within colorectal cancer specimens. This classification is particularly relevant because current studies have illustrated that these distinctions can have profound implications on treatment choices and patient prognoses. As such, the study underscores not only the accuracy of automated assessments but also their potential impact on clinical outcomes for patients receiving treatment for colorectal cancer.

Moreover, the alignment between automated calculations performed by the machine learning model and manual assessments highlights a breakthrough in ensuring that technology complements, rather than competes with, human expertise. The ability of AI systems to achieve such a strong correlation indicates their readiness for adoption into standard pathological practices, paving the way for more scalable and standardized approaches to cancer diagnosis.

The implications of employing a hybrid deep learning framework extend beyond colorectal cancer. As research in this arena develops, the methodology has the potential to be adapted for other cancer types, representing a significant advancement in the overarching strategy employed in oncological diagnostics. This adaptability emphasizes the versatility and robustness of deep learning systems, preparing them for broader application in various domains of cancer care.

With a focus on integrating these advanced systems into existing pathological workflows, the research addresses the urgent need for solutions that enhance diagnostic accuracy while also alleviating the workload burden on pathologists. As diagnostic cases continue to increase worldwide, the role of AI becomes ever more critical in ensuring that clinicians can maintain high standards of care without being overwhelmed.

The practical benefits of utilizing hybrid deep learning systems are manifold. Not only do they promise quicker turnaround times for diagnostic decisions, but they also aim to reduce subjective variability that can occur when assessments are conducted manually. This aspect is particularly vital when considering that patient outcomes can hinge upon the clarity and accuracy of such assessments. In this light, the evolution towards digital pathology, powered by AI technology, appears both timely and necessary.

As the research unfolds, it becomes evident that the potential for machine learning approaches in the realm of oncology is expansive. By accelerating the process of pathological evaluation, they represent a forward-thinking strategy to overcome the hurdles posed by traditional diagnostic methodologies. The aim is not merely to replace human pathologists but to create an ecosystem where technology augments human analysis, achieving a new zenith in medical diagnostics.

The journey of integrating advanced deep learning frameworks into clinical routine is still in its early stages. However, the promising results presented by the Efficient-TransUNet introduce a paradigm characterized by greater accuracy, heightened efficiency, and improved outcomes for patients confronting the challenges of colorectal cancer. The roadmap ahead encourages further exploration, expecting even more breakthroughs as the synergy between technology and medicine deepens.

Thus, the research not only provides a glimpse into the future of cancer diagnostics but also ignites hope for improved therapeutic strategies that can significantly enhance the quality of life for patients affected by colorectal cancer. In a world where technology continues to reshape various facets of life, its convergence with healthcare indicates a promising frontier worth watching as we stride into a new age of medical innovation.

Subject of Research: Tumor-Stroma Ratio (TSR) analysis in colorectal cancer using deep learning

Article Title: Automated tumor stroma ratio assessment in colorectal cancer using hybrid deep learning approach.

Article References:

Armand, T.P.T., Bhattacharjee, S., Nfor, K.A. et al. Automated tumor stroma ratio assessment in colorectal cancer using hybrid deep learning approach.
Sci Rep 15, 40927 (2025). https://doi.org/10.1038/s41598-025-24229-8

Image Credits: AI Generated

DOI: https://doi.org/10.1038/s41598-025-24229-8

Keywords: Deep learning, colorectal cancer, tumor-stroma ratio, convolutional neural networks, transformers, histopathology, automated assessment