Elamathi Natarajan
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Elamathi Natarajan has a total of 17 co-authors. Here are the top co-authors by number of shared publications:
- Akshata P. Pauskar
Department of Bioinformatics, AI & Machine Learning Biotecnika Info Labs Pvt Ltd, Bangalore, India
2 shared publications - Subhasankar Khilar
Biotecnika
2 shared publications - Aby John
Bioinformatics, Biotecnika Info Labs Pvt. Ltd., Bengaluru, Karnataka, India
1 shared publication - Aman Zehra
Department of Bioinformatics, AI & Machine Learning, Biotecnika Info Labs, Bangalore, India
1 shared publication - Aryan Anand
Department of Bioinformatics, Biotecnika Info Labs Pvt Ltd, Bengaluru, India
1 shared publication - Dudyala Sri Lakshmi
Biotecnika, HSR Layout, Bengaluru
1 shared publication
Latest Publications
- Open Access
Integrative Multiomics Analysis of the Angiogenesis–EMT Interface in Head and Neck Squamous Cell Carcinoma
Angiogenesis and epithelial–mesenchymal transition (EMT) are key hallmarks that contribute to the progression of oral squamous cell carcinoma (OSCC), but whether the two programs vary simultaneously across tumors and how much of this reflects tumor composition remain unclear. We scored predefined, non-overlapping 20-gene angiogenesis and EMT signatures with a rank-based method in 342 primary oral cavity tumors from The Cancer Genome Atlas (TCGA-OSCC). The two programs were strongly correlated (Spearman ρ = 0.809, P = 2.7 × 10 −80 ), and 2,396 bridge genes were associated with both programs across the transcriptome. The bridge genes were enriched in pathways associated with focal adhesion and extracellular matrix– receptor interaction. In the cellular case, fibroblast and endothelial marker scores were prominent and correlated with the program scores. The coupling was attenuated but remained positive after joint adjustment for tumor purity and stromal and immune scores (partial ρ = 0.406) and with an epithelial-restricted EMT score (ρ = 0.710). Further correlation was performed with a partial EMT (pEMT), which shows a significant positive correlation (ρ = 0.59) and an inverse correlation with tumor purity (ρ = −0.70). 146 final gene candidates were found using multi-omic evidence and candidate-level testing on the external cohort GSE30784; after adjusting for inferred cell fractions, 101 of these candidates supported the dual program. For thorough characterization, three candidates with an emphasis on OSMR, MICAL2, and LHFPL2 were chosen. Additional CPTAC data were used to confirm protein-level support, which was strongest for MICAL2. No significant connection was seen between the Interface score and overall survival (hazard ratio per 1-SD increase, 0.92; 95% CI, 0.78–1.08). These results describe a coordinated angiogenesis–EMT transcriptional state in OSCC that is attenuated but retained after accounting for tumor composition, and nominate candidate genes for experimental study.
Open AccessIntegrative Multiomics Analysis of the Angiogenesis–EMT Interface in Head and Neck Squamous Cell Carcinoma
Angiogenesis and epithelial–mesenchymal transition (EMT) are key hallmarks that contribute to the progression of oral squamous cell carcinoma (OSCC), but whether the two programs vary simultaneously across tumors and how much of this reflects tumor composition remain unclear. We scored predefined, non-overlapping 20-gene angiogenesis and EMT signatures with a rank-based method in 342 primary oral cavity tumors from The Cancer Genome Atlas (TCGA-OSCC). The two programs were strongly correlated (Spearman ρ = 0.809, P = 2.7 × 10 −80 ), and 2,396 bridge genes were associated with both programs across the transcriptome. The bridge genes were enriched in pathways associated with focal adhesion and extracellular matrix– receptor interaction. In the cellular case, fibroblast and endothelial marker scores were prominent and correlated with the program scores. The coupling was attenuated but remained positive after joint adjustment for tumor purity and stromal and immune scores (partial ρ = 0.406) and with an epithelial-restricted EMT score (ρ = 0.710). Further correlation was performed with a partial EMT (pEMT), which shows a significant positive correlation (ρ = 0.59) and an inverse correlation with tumor purity (ρ = −0.70). 146 final gene candidates were found using multi-omic evidence and candidate-level testing on the external cohort GSE30784; after adjusting for inferred cell fractions, 101 of these candidates supported the dual program. For thorough characterization, three candidates with an emphasis on OSMR, MICAL2, and LHFPL2 were chosen. Additional CPTAC data were used to confirm protein-level support, which was strongest for MICAL2. No significant connection was seen between the Interface score and overall survival (hazard ratio per 1-SD increase, 0.92; 95% CI, 0.78–1.08). These results describe a coordinated angiogenesis–EMT transcriptional state in OSCC that is attenuated but retained after accounting for tumor composition, and nominate candidate genes for experimental study.
- 2 November 2026
- Published by Springer Nature in In Silico Pharmacology
- Vol. 14 (3) , 276
Computational identification of EBNA1-binding candidates using machine learning-based virtual screening and molecular docking
- 2 November 2026
- Published by Springer Nature in In Silico Pharmacology
- Vol. 14 (3) , 276
Computational identification of EBNA1-binding candidates using machine learning-based virtual screening and molecular docking
- Open Access
- 1 November 2026
- Published by GSC Online Press in World Journal of Advanced Research and Reviews
- Vol. 32 (2) , 1314
PREDICTING HIV DRUG RESISTANCE MUTATIONS: BIOINFORMATIC STRATEGIES, MACHINE-LEARNING MODELS, AND CLINICAL RELEVANCE
HIV-1 is characterised by a high degree of genetic diversity, and it is this diversity that leads to resistance against antiretroviral drugs. Resistance occurs when changes in the viral genome reduce the virus's sensitivity to one or more of the antiretroviral agents. By identifying and predicting the mutations associated with resistance, clinicians and researchers are able to explain why treatment fails, monitor the way in which the virus evolves, and select appropriate treatment regimens. Traditional genotypic interpretation makes use of established mutation patterns and scoring systems in order to estimate susceptibility. More recently, progress in the fields of bioinformatics and machine learning has expanded the range of methods available for deducing resistance from viral sequences, the methods extending from statistical regression and support vector machines to random forests, gradient boosting and deep neural networks. Resources that are openly available, especially the Stanford HIV Drug Resistance Database, provide paired genotype and phenotype records which enable model building and benchmarking. Yet, despite this, it is still very difficult to make reliable predictions, due to the diversity of viruses, the effect of interacting mutations, imbalanced class distributions, inconsistent definitions of resistance, and the lack of independently validated data. This review addresses the biology of HIV drug resistance, the existing interpretation tools, computational prediction methods, evaluation practices, and the open problems, and looks at how interpretable and widely applicable systems might be developed. The use of computational models together with established interpretation systems and appropriate clinical evidence could make resistance assessment more reliable.
- 1 November 2026
- Published by GSC Online Press in World Journal of Advanced Research and Reviews
- Vol. 32 (2) , 1314
Open AccessPREDICTING HIV DRUG RESISTANCE MUTATIONS: BIOINFORMATIC STRATEGIES, MACHINE-LEARNING MODELS, AND CLINICAL RELEVANCE
HIV-1 is characterised by a high degree of genetic diversity, and it is this diversity that leads to resistance against antiretroviral drugs. Resistance occurs when changes in the viral genome reduce the virus's sensitivity to one or more of the antiretroviral agents. By identifying and predicting the mutations associated with resistance, clinicians and researchers are able to explain why treatment fails, monitor the way in which the virus evolves, and select appropriate treatment regimens. Traditional genotypic interpretation makes use of established mutation patterns and scoring systems in order to estimate susceptibility. More recently, progress in the fields of bioinformatics and machine learning has expanded the range of methods available for deducing resistance from viral sequences, the methods extending from statistical regression and support vector machines to random forests, gradient boosting and deep neural networks. Resources that are openly available, especially the Stanford HIV Drug Resistance Database, provide paired genotype and phenotype records which enable model building and benchmarking. Yet, despite this, it is still very difficult to make reliable predictions, due to the diversity of viruses, the effect of interacting mutations, imbalanced class distributions, inconsistent definitions of resistance, and the lack of independently validated data. This review addresses the biology of HIV drug resistance, the existing interpretation tools, computational prediction methods, evaluation practices, and the open problems, and looks at how interpretable and widely applicable systems might be developed. The use of computational models together with established interpretation systems and appropriate clinical evidence could make resistance assessment more reliable.
- Open Access
- 1 November 2026
- Published by GSC Online Press in World Journal of Advanced Research and Reviews
- Vol. 32 (2) , 982
COMPUTATIONAL AND MACHINE LEARNING APPROACHES FOR SINGLE-CELL TRANSCRIPTOMICS: FROM CELLULAR HETEROGENEITY TO BIOLOGICAL DISCOVERY
Single-cell RNA sequencing (scRNA-seq) has transformed gene expression analysis by revealing cellular heterogeneity that is obscured in bulk tissue measurements. This is particularly relevant to lung adenocarcinoma (LUAD), where malignant, immune, stromal, and endothelial cells interact within a complex tumor microenvironment. However, the high dimensionality, sparsity, sequencing variability, and batch effects of scRNA-seq data present substantial challenges for reliable biological interpretation. This review examines computational and machine-learning approaches across the scRNA-seq analysis workflow, encompassing quality control, normalization, feature selection, dimensionality reduction, clustering, and cell annotation. Conventional methods, including k-means, hierarchical, and graph-based clustering, are considered alongside advanced approaches such as autoencoders, variational autoencoders, and graph neural networks. Their applications in downstream analyses, including differential expression, pathway enrichment, trajectory inference, and cell-cell communication, are discussed within a LUAD-focused context. The review further evaluates how analytical assumptions, model interpretability, and validation practices influence biological conclusions. Emerging developments in spatial transcriptomics, multimodal integration, and explainable machine learning are also highlighted. The key contribution is an integrated, validation-oriented framework connecting computational outputs with biological interpretation and evidence-supported claims, moving beyond a descriptive catalogue of methods. This perspective emphasizes rigorous method selection and independent validation to strengthen insights into cellular heterogeneity and LUAD biology.
- 1 November 2026
- Published by GSC Online Press in World Journal of Advanced Research and Reviews
- Vol. 32 (2) , 982
Open AccessCOMPUTATIONAL AND MACHINE LEARNING APPROACHES FOR SINGLE-CELL TRANSCRIPTOMICS: FROM CELLULAR HETEROGENEITY TO BIOLOGICAL DISCOVERY
Single-cell RNA sequencing (scRNA-seq) has transformed gene expression analysis by revealing cellular heterogeneity that is obscured in bulk tissue measurements. This is particularly relevant to lung adenocarcinoma (LUAD), where malignant, immune, stromal, and endothelial cells interact within a complex tumor microenvironment. However, the high dimensionality, sparsity, sequencing variability, and batch effects of scRNA-seq data present substantial challenges for reliable biological interpretation. This review examines computational and machine-learning approaches across the scRNA-seq analysis workflow, encompassing quality control, normalization, feature selection, dimensionality reduction, clustering, and cell annotation. Conventional methods, including k-means, hierarchical, and graph-based clustering, are considered alongside advanced approaches such as autoencoders, variational autoencoders, and graph neural networks. Their applications in downstream analyses, including differential expression, pathway enrichment, trajectory inference, and cell-cell communication, are discussed within a LUAD-focused context. The review further evaluates how analytical assumptions, model interpretability, and validation practices influence biological conclusions. Emerging developments in spatial transcriptomics, multimodal integration, and explainable machine learning are also highlighted. The key contribution is an integrated, validation-oriented framework connecting computational outputs with biological interpretation and evidence-supported claims, moving beyond a descriptive catalogue of methods. This perspective emphasizes rigorous method selection and independent validation to strengthen insights into cellular heterogeneity and LUAD biology.
- Open Access
Integrated 16S rRNA Bioinformatics Analysis of Gut Microbiota, Predicted Pathways, and Candidate Biomarkers in Inflammatory Bowel Disease
Inflammatory bowel disease (IBD) is associated with changes in gut microbiota composition and predicted microbial functions that may relate to intestinal immune processes. This study compared gut microbiota profiles between individuals with IBD and healthy controls and examined predicted microbial pathways and their associations with bacterial genera. Publicly available 16S rRNA amplicon sequencing data from 22 stool samples (11 IBD and 11 healthy controls) were analysed. Taxonomic profiling, diversity analysis, differential abundance testing, and PICRUSt2-based functional prediction were performed. Microbe–pathway associations were examined using Spearman correlation and Cytoscape. Extended analyses integrated differential abundance with machine-learning feature importance to prioritize candidate genera, evaluated a three-genus signature, assessed selected findings in an independent dataset, and compared predicted fermentation- and acetate-related pathways. Alpha diversity was significantly lower in the IBD group (Shannon index, p = 0.003), and Bray–Curtis PCoA showed separation of the two groups mainly along PC1 (PERMANOVA pseudo-F = 6.125, p < 0.001), although this separation may partly reflect differences between the source cohorts. DESeq2 identified 44 significantly differentially abundant genera among 186 tested (FDR-adjusted p < 0.05), of which 32 were lower and 12 were higher in the IBD group relative to healthy controls; 27 of these genera were detected in only one of the two groups. PICRUSt2 predicted lipopolysaccharide and peptidoglycan biosynthesis pathways; after normalising pathway abundances for sequencing depth, Spearman analysis identified 84 significant microbe–pathway associations, 47 with lipopolysaccharide biosynthesis (44 positive; ρ up to 0.90) and 37 with peptidoglycan biosynthesis (33 negative; ρ down to −0.82). Five genera were prioritized by integrated biomarker analysis. The combined three-genus signature achieved an ROC–AUC of 0.959. External validation was limited to Bifidobacterium, whose abundance direction was concordant with the discovery analysis. Of the two SCFA-related pathways examined, the predicted acetate-related pathway (methanogenesis from acetate) was lower in the IBD group (FDR = 0.002), whereas the predicted fermentation pathway did not differ significantly between groups (FDR = 0.076). These findings provide an integrated view of microbial composition and predicted functional potential in IBD. The small sample size, differences in cohort and sequencing characteristics between the IBD and control samples, limited external validation, and reliance on predicted rather than directly measured functions mean that the results are exploratory and require confirmation in larger independent cohorts and through direct functional or metabolomic analyses.
Open AccessIntegrated 16S rRNA Bioinformatics Analysis of Gut Microbiota, Predicted Pathways, and Candidate Biomarkers in Inflammatory Bowel Disease
Inflammatory bowel disease (IBD) is associated with changes in gut microbiota composition and predicted microbial functions that may relate to intestinal immune processes. This study compared gut microbiota profiles between individuals with IBD and healthy controls and examined predicted microbial pathways and their associations with bacterial genera. Publicly available 16S rRNA amplicon sequencing data from 22 stool samples (11 IBD and 11 healthy controls) were analysed. Taxonomic profiling, diversity analysis, differential abundance testing, and PICRUSt2-based functional prediction were performed. Microbe–pathway associations were examined using Spearman correlation and Cytoscape. Extended analyses integrated differential abundance with machine-learning feature importance to prioritize candidate genera, evaluated a three-genus signature, assessed selected findings in an independent dataset, and compared predicted fermentation- and acetate-related pathways. Alpha diversity was significantly lower in the IBD group (Shannon index, p = 0.003), and Bray–Curtis PCoA showed separation of the two groups mainly along PC1 (PERMANOVA pseudo-F = 6.125, p < 0.001), although this separation may partly reflect differences between the source cohorts. DESeq2 identified 44 significantly differentially abundant genera among 186 tested (FDR-adjusted p < 0.05), of which 32 were lower and 12 were higher in the IBD group relative to healthy controls; 27 of these genera were detected in only one of the two groups. PICRUSt2 predicted lipopolysaccharide and peptidoglycan biosynthesis pathways; after normalising pathway abundances for sequencing depth, Spearman analysis identified 84 significant microbe–pathway associations, 47 with lipopolysaccharide biosynthesis (44 positive; ρ up to 0.90) and 37 with peptidoglycan biosynthesis (33 negative; ρ down to −0.82). Five genera were prioritized by integrated biomarker analysis. The combined three-genus signature achieved an ROC–AUC of 0.959. External validation was limited to Bifidobacterium, whose abundance direction was concordant with the discovery analysis. Of the two SCFA-related pathways examined, the predicted acetate-related pathway (methanogenesis from acetate) was lower in the IBD group (FDR = 0.002), whereas the predicted fermentation pathway did not differ significantly between groups (FDR = 0.076). These findings provide an integrated view of microbial composition and predicted functional potential in IBD. The small sample size, differences in cohort and sequencing characteristics between the IBD and control samples, limited external validation, and reliance on predicted rather than directly measured functions mean that the results are exploratory and require confirmation in larger independent cohorts and through direct functional or metabolomic analyses.
- Open Access
- 17 October 2026
- Published by IORO Publications
- Vol. 1 (1) , 57-72
Single-cell RNA sequencing of the Tumor Microenvironment in Gynaecological Cancer: Cellular Heterogeneity, Computational Approaches and Clinical Applications
Single-cell RNA sequencing (scRNA-seq) has emerged as a powerful technological advancement, overcoming the primary limitation of bulk-RNA sequencing–its inability to capture cellular heterogeneity. Consequently, scRNA-seq plays a pivotal role in characterising the complex cellular landscape of the tumor microenvironment. Composed of diverse populations, including immune and stromal cells, and characterised by intricate cell-to-cell interactions, the tumor microenvironment is a central driver of tumorigenesis. Examining these dynamic biological systems at the transcriptomic level enables detailed mapping of intercellular communications and facilitates the discovery of targeted therapeutic strategies. This review outlines the end-to-end scRNA-seq workflow, covering fundamental single-cell isolation methods, downstream computational analyses, and key analytical software, as well as their practical significance. Additionally, it highlights the integration of spatial transcriptomics, which provides crucial positional context to achieve a more comprehensive understanding of intra-tumoral heterogeneity. Special emphasis is placed on gynaecological malignancies, particularly cervical and endometrial cancers. Given the rising incidence of both malignancies in recent years, detailed characterisation of their specific microenvironments is critical for advancing diagnostic accuracy and therapeutic intervention. A comprehensive understanding of the progression of the cancer and drug resistance is critical to understand. Accordingly, key molecular alterations, potential gene targets, and emerging biomarkers are systematically reviewed. Finally, current technical constraints, recent technological advancements, and future perspectives for single-cell transcriptomics are evaluated. Keywords: Keywords: Cell-Cell communication; Cervical cancer; Endometrial cancer; Spatial transcriptomics; Tools
- 17 October 2026
- Published by IORO Publications
- Vol. 1 (1) , 57-72
Open AccessSingle-cell RNA sequencing of the Tumor Microenvironment in Gynaecological Cancer: Cellular Heterogeneity, Computational Approaches and Clinical Applications
Single-cell RNA sequencing (scRNA-seq) has emerged as a powerful technological advancement, overcoming the primary limitation of bulk-RNA sequencing–its inability to capture cellular heterogeneity. Consequently, scRNA-seq plays a pivotal role in characterising the complex cellular landscape of the tumor microenvironment. Composed of diverse populations, including immune and stromal cells, and characterised by intricate cell-to-cell interactions, the tumor microenvironment is a central driver of tumorigenesis. Examining these dynamic biological systems at the transcriptomic level enables detailed mapping of intercellular communications and facilitates the discovery of targeted therapeutic strategies. This review outlines the end-to-end scRNA-seq workflow, covering fundamental single-cell isolation methods, downstream computational analyses, and key analytical software, as well as their practical significance. Additionally, it highlights the integration of spatial transcriptomics, which provides crucial positional context to achieve a more comprehensive understanding of intra-tumoral heterogeneity. Special emphasis is placed on gynaecological malignancies, particularly cervical and endometrial cancers. Given the rising incidence of both malignancies in recent years, detailed characterisation of their specific microenvironments is critical for advancing diagnostic accuracy and therapeutic intervention. A comprehensive understanding of the progression of the cancer and drug resistance is critical to understand. Accordingly, key molecular alterations, potential gene targets, and emerging biomarkers are systematically reviewed. Finally, current technical constraints, recent technological advancements, and future perspectives for single-cell transcriptomics are evaluated. Keywords: Keywords: Cell-Cell communication; Cervical cancer; Endometrial cancer; Spatial transcriptomics; Tools
- Open Access
Cross-Dataset Reanalysis of Mouse Alzheimer Disease Single-Cell RNA-Sequencing Data Reveals a Reproducible Innate Immune Activation Signature
Single-cell RNA sequencing (scRNA-seq) studies of Alzheimer’s disease (AD) mouse models have proliferated across independent laboratories, brain regions, and platforms, yet the cross-study reproducibility of transcriptional signatures derived from such data remains underexplored. We reanalyzed six publicly available mouse brain scRNA-seq datasets (GSE140399, GSE140510, GSE143758, GSE208683, GSE214244, and GSE296027; 360,620 cells, 102 annotated clusters) using a standardized pipeline (Figure 1) comprising quality control, Leiden clustering, confidence-tiered marker-based cell-type annotation, sample-level pseudobulk differential expression, cross-dataset consistency analysis, and pathway enrichment. Pseudobulk differential expression, performed on 32 samples (16 AD, 16 Control) from five of the six datasets, identified 301 significant genes (FDR-adjusted p < 0.05) out of 20,629 tested, of which 290 (96.3%) were upregulated and 11 (3.7%) downregulated in AD. The upregulated set was dominated by innate immune and interferon-response genes (Ccl3, Ccl4, Ccl6, Itgax, Cst7, Tlr4, Trem2) and showed extremely strong enrichment for immune system and innate immune system pathways (Reactome, p = 1.9 × 10□□³ and 2.1 × 10□²□ respectively), phagosome and osteoclast differentiation (KEGG), and interferon alpha/beta signaling. Of the 301 significant genes, 235 (78%) showed the same direction of effect across all datasets in which they were tested. The sixth dataset, GSE296027, was excluded from differential expression analysis due to a gene-identifier corruption artifact introduced during pseudobulk aggregation (detailed in Methods and Limitations) but contributed valid cell-type composition and microglial subclustering data, independently recovering a disease-associated microglia-like state. These findings indicate that a reproducible, strongly immune/interferon-skewed transcriptional signature, rather than a balanced up/down transcriptional shift, characterizes the mouse AD models analyzed here.
Open AccessCross-Dataset Reanalysis of Mouse Alzheimer Disease Single-Cell RNA-Sequencing Data Reveals a Reproducible Innate Immune Activation Signature
Single-cell RNA sequencing (scRNA-seq) studies of Alzheimer’s disease (AD) mouse models have proliferated across independent laboratories, brain regions, and platforms, yet the cross-study reproducibility of transcriptional signatures derived from such data remains underexplored. We reanalyzed six publicly available mouse brain scRNA-seq datasets (GSE140399, GSE140510, GSE143758, GSE208683, GSE214244, and GSE296027; 360,620 cells, 102 annotated clusters) using a standardized pipeline (Figure 1) comprising quality control, Leiden clustering, confidence-tiered marker-based cell-type annotation, sample-level pseudobulk differential expression, cross-dataset consistency analysis, and pathway enrichment. Pseudobulk differential expression, performed on 32 samples (16 AD, 16 Control) from five of the six datasets, identified 301 significant genes (FDR-adjusted p < 0.05) out of 20,629 tested, of which 290 (96.3%) were upregulated and 11 (3.7%) downregulated in AD. The upregulated set was dominated by innate immune and interferon-response genes (Ccl3, Ccl4, Ccl6, Itgax, Cst7, Tlr4, Trem2) and showed extremely strong enrichment for immune system and innate immune system pathways (Reactome, p = 1.9 × 10□□³ and 2.1 × 10□²□ respectively), phagosome and osteoclast differentiation (KEGG), and interferon alpha/beta signaling. Of the 301 significant genes, 235 (78%) showed the same direction of effect across all datasets in which they were tested. The sixth dataset, GSE296027, was excluded from differential expression analysis due to a gene-identifier corruption artifact introduced during pseudobulk aggregation (detailed in Methods and Limitations) but contributed valid cell-type composition and microglial subclustering data, independently recovering a disease-associated microglia-like state. These findings indicate that a reproducible, strongly immune/interferon-skewed transcriptional signature, rather than a balanced up/down transcriptional shift, characterizes the mouse AD models analyzed here.
- Open Access
- 1 October 2026
- Published by GSC Online Press in World Journal of Advanced Research and Reviews
- Vol. 32 (1) , 1714-1727
UNRAVELING TUMOR HETEROGENEITY AND PROGNOSTIC BIOMARKERS IN GLIOBLASTOMA THROUGH INTEGRATIVE EXOMIC PROFILING AND SUPERVISED CLASSIFIERS: A PILOT TCGA COHORT ANALYSIS
Glioblastoma Multiforme (GBM) represents the most prevalent and highly aggressive primary malignant tumor of the brain, exhibiting substantial molecular heterogeneity, intricate signaling pathways, and poor clinical outcomes, high-with a median survival of approximately 12-18 months even following standard trimodal treatment. Conventional histopathological classification and single-gene-based clinical strategies are often insufficient to fully characterize the complex and multidimensional genomic architecture of GBM, emphasizing the importance of precision medicine approaches. This review and integrated computational study explore the potential of combing Next-Generation sequencing (NGS), particularly Whole-Exome Sequencing (WES), with Artificial Intelligence (AI) and Machine (ML) methodologies for systematic analysis of high-dimensional genomic data, reduction of background variation, and identification of clinically relevant pathogenic somatic driver mutations, including alterations in TP53, PTEN, EGFR, and IDH1. Clinical and genomic information from the Cancer Genome Atlas (TCGA-GBM; N=79) was integrated to analyze feature-selection strategies, principal component analysis (PCA), Kaplan-Meier survival analysis based on molecular biomarker status such as PTEN and regularized logistic regression models designed to address class imbalance and high-dimensional genomic sparsity. In addition, the study considers the influence of the tumor microenvironment, epigenetic factors such as MGMT promoter methylation, and emerging evidence of sex-associated molecular differences in GBM. Collectively, the findings emphasize the potential of interpretable and integrative multi-omics-based machine learning frameworks to improve molecular and prognostic characterization of GBM and support the development of more precise, biomarker-guided therapeutic strategies in neuro-oncology.
- 1 October 2026
- Published by GSC Online Press in World Journal of Advanced Research and Reviews
- Vol. 32 (1) , 1714-1727
Open AccessUNRAVELING TUMOR HETEROGENEITY AND PROGNOSTIC BIOMARKERS IN GLIOBLASTOMA THROUGH INTEGRATIVE EXOMIC PROFILING AND SUPERVISED CLASSIFIERS: A PILOT TCGA COHORT ANALYSIS
Glioblastoma Multiforme (GBM) represents the most prevalent and highly aggressive primary malignant tumor of the brain, exhibiting substantial molecular heterogeneity, intricate signaling pathways, and poor clinical outcomes, high-with a median survival of approximately 12-18 months even following standard trimodal treatment. Conventional histopathological classification and single-gene-based clinical strategies are often insufficient to fully characterize the complex and multidimensional genomic architecture of GBM, emphasizing the importance of precision medicine approaches. This review and integrated computational study explore the potential of combing Next-Generation sequencing (NGS), particularly Whole-Exome Sequencing (WES), with Artificial Intelligence (AI) and Machine (ML) methodologies for systematic analysis of high-dimensional genomic data, reduction of background variation, and identification of clinically relevant pathogenic somatic driver mutations, including alterations in TP53, PTEN, EGFR, and IDH1. Clinical and genomic information from the Cancer Genome Atlas (TCGA-GBM; N=79) was integrated to analyze feature-selection strategies, principal component analysis (PCA), Kaplan-Meier survival analysis based on molecular biomarker status such as PTEN and regularized logistic regression models designed to address class imbalance and high-dimensional genomic sparsity. In addition, the study considers the influence of the tumor microenvironment, epigenetic factors such as MGMT promoter methylation, and emerging evidence of sex-associated molecular differences in GBM. Collectively, the findings emphasize the potential of interpretable and integrative multi-omics-based machine learning frameworks to improve molecular and prognostic characterization of GBM and support the development of more precise, biomarker-guided therapeutic strategies in neuro-oncology.
- Open Access
- 6 August 2026
- Published by Springer Nature in Research Square
Spatial GNN Analyzer: An AI-Powered Graph Neural Network Framework for Spatial Transcriptomics Domain Analysis and Visualization
Spatial transcriptomics has developed as a game-changer technology for understanding tissue organization by converting them into spatial context of gene expression. Due to the high dimensionality & complex spatial dependencies within transcriptomics, that is challenging to accurately identify spatial domains and uncover biologically meaningful tissue structures. This work presents Spatial GNN Analyzer, which is an AI-powered analytical frame-work it integrated with Graph Neural Networks (GNNs) with spatial transcriptomics to enable robust spatial domain identification, interactive tissue analysis, representation learning & an end-to-end pipeline. Here, my proposed framework models spatial transcriptomics data as a graph representation in which all the tissue spots are represented as nodes whereas the spatial neighbourhood relationships are encoded as graph edges. The system learns both local cellular interactions and global tissue architecture which are biologically informative latent embeddings capable of capturing by combining transcriptomic profiles with spatial topology through Graph Convolutional Networks (GCNs). The pipeline works a comprehensive preprocessing steps like quality control, normalization, highly variable gene selectin, dimensionality reduction, & spatial graph construction followed by graph-based clustering & domain segmentation. It utilizes Scanpy & Squidpy for preprocessing & spatial graph generation, followed by graph-aware clustering & visualization. By using Leiden clustering & UMAP embedding the spatial coherent tissue was identified and visualization was done respectably. To figure out clustering performance & biological coherence, the framework integrated metrics such as Silhouette score & Spatial smoothness, to ensure transcriptomics separability & spatial continuity of predicted domains. The Silhouette score is 0.37 which indicates effective preservation of local tissue structure & the Spatial smoothness score is 0.93, which shows a strong spatial coherence and meaningful cluster separation of Visium breast cancer spatial transcriptomics. An interactive visualization platform was developed to do exploration of spatial clusters, tissue morphology, & gene expression patterns through dynamic multi-panel visual analytical and another best thing is automated report generation facility. This framework was validated using publicly available 10x Genomics Visium datasets, including breast cancer spatial transcriptomics samples, that shows the improved preservation of tissue architecture & enhanced spatially coherent clustering compared with conventional non-spatial analytical approaches. This work will provide a scalable & interpretable computational solution for next-generation spatial biology research as it integrated with deep graph learning, spatial omics analysis, & an interactive visualization into a uniform platform. The proposed work has a broad application in tumour microenvironment analysis, biomarker discovery, tissue heterogeneity characterization, & precision medicine.
- 6 August 2026
- Published by Springer Nature in Research Square
Open AccessSpatial GNN Analyzer: An AI-Powered Graph Neural Network Framework for Spatial Transcriptomics Domain Analysis and Visualization
Spatial transcriptomics has developed as a game-changer technology for understanding tissue organization by converting them into spatial context of gene expression. Due to the high dimensionality & complex spatial dependencies within transcriptomics, that is challenging to accurately identify spatial domains and uncover biologically meaningful tissue structures. This work presents Spatial GNN Analyzer, which is an AI-powered analytical frame-work it integrated with Graph Neural Networks (GNNs) with spatial transcriptomics to enable robust spatial domain identification, interactive tissue analysis, representation learning & an end-to-end pipeline. Here, my proposed framework models spatial transcriptomics data as a graph representation in which all the tissue spots are represented as nodes whereas the spatial neighbourhood relationships are encoded as graph edges. The system learns both local cellular interactions and global tissue architecture which are biologically informative latent embeddings capable of capturing by combining transcriptomic profiles with spatial topology through Graph Convolutional Networks (GCNs). The pipeline works a comprehensive preprocessing steps like quality control, normalization, highly variable gene selectin, dimensionality reduction, & spatial graph construction followed by graph-based clustering & domain segmentation. It utilizes Scanpy & Squidpy for preprocessing & spatial graph generation, followed by graph-aware clustering & visualization. By using Leiden clustering & UMAP embedding the spatial coherent tissue was identified and visualization was done respectably. To figure out clustering performance & biological coherence, the framework integrated metrics such as Silhouette score & Spatial smoothness, to ensure transcriptomics separability & spatial continuity of predicted domains. The Silhouette score is 0.37 which indicates effective preservation of local tissue structure & the Spatial smoothness score is 0.93, which shows a strong spatial coherence and meaningful cluster separation of Visium breast cancer spatial transcriptomics. An interactive visualization platform was developed to do exploration of spatial clusters, tissue morphology, & gene expression patterns through dynamic multi-panel visual analytical and another best thing is automated report generation facility. This framework was validated using publicly available 10x Genomics Visium datasets, including breast cancer spatial transcriptomics samples, that shows the improved preservation of tissue architecture & enhanced spatially coherent clustering compared with conventional non-spatial analytical approaches. This work will provide a scalable & interpretable computational solution for next-generation spatial biology research as it integrated with deep graph learning, spatial omics analysis, & an interactive visualization into a uniform platform. The proposed work has a broad application in tumour microenvironment analysis, biomarker discovery, tissue heterogeneity characterization, & precision medicine.
- Open Access
Deciphering the Tumor Microenvironment: An Integrated Single-Cell RNA-Seq and AI Framework for Novel Biomarker and Therapeutic Target Discovery in Melanoma
Background: Melanoma represents a highly immunogenic and therapeutically challenging malignancy. The complex cellular ecosystem of the tumor microenvironment (TME) acts as a critical driver of immune evasion, patient prognosis, and treatment resistance. Traditional bulk sequencing often fails to resolve these high-resolution intercellular dynamics. This investigation presents an end-to-end explainable machine learning and network biology framework designed to deconstruct TME single-cell heterogeneity, discover novel candidate biomarkers, and map actionable cell-cell communication networks. Methodology: High-quality single-cell RNA sequencing (scRNA-seq) expression data from melanoma lesions (GSE115978) were processed using a multi-phase computational workflow. Following cellular filtration, library size normalisation, and highly variable gene (HVG) selection, cells were partitioned using unsupervised Leiden clustering and annotated via signature gene scoring matrix methods (Wolf et al., 2018; Traag et al., 2019). An optimal gradient-boosted tree ensemble classifier (XGBoost) was constructed for cross-compartment biomarker screening (Chen & Guestrin, 2016), interpreted using SHAP (SHapley Additive exPlanations) values (Lundberg & Lee, 2017), and augmented with a PyTorch deep autoencoder. Downstream systems analyses included biological pathway enrichment via gseapy (Kuleshov et al., 2016), ligand-receptor communication mapping via LIANA (Efremova et al., 2020; Dimitrov et al., 2022), and Pearson co-expression network profiling. Cross-cohort validation was conducted on an independent melanoma cohort (GSE72056) (Tirosh et al., 2016b), and prognostic utility was clinically validated using empirical patient survival data from the TCGA-SKCM cohort (TCGA Research Network, 2015; Davidson-Pilon, 2019). Results: Unsupervised Leiden clustering partitioned the cellular atlas into seven major structural and immunological compartments: Melanoma/Tumor, T-Cells, B-Cells, Macrophages, NK Cells, Endothelial cells, and Cancer-Associated Fibroblasts (CAFs). The trained XGBoost model prioritized CD79A, MLANA, LYZ, MFAP4 , and CDH5 as top candidate biomarkers across distinct microenvironmental niches. SHAP explainability analysis confirmed MLANA and S100B as key positive predictors of tumor cell identity, while highlighting a bidirectional distribution for B2M linked to antigen presentation downregulation. Functional enrichment mapped Coagulation and Epithelial-Mesenchymal Transition (EMT) as highly dysregulated pathways driving the malignant state. Cell-cell communication profiling inferred highly significant SERPINE1 signalling axes to LRP1 and PLAUR receptors. SPARC was identified as the central co-expression hub regulator. Cross-cohort screening in GSE72056 confirmed biomarker stability across independent patient profiles. Empirical clinical survival validation within the TCGA-SKCM cohort (n=314) demonstrated that heightened expression of CD79A correlates with a statistically significant survival advantage (Log-Rank p = 4.4202 × 10-3), extending median overall survival from 26.7 months to 44.8 months. Conclusion: This computational pipeline systematically resolves TME heterogeneity, revealing a robust biomarker signature centred on CD79A and MLANA, alongside SERPINE1-driven immune-stromal crosstalk. The discovery of the protective prognostic role of CD79A links single-cell immune networks directly to clinical patient outcomes, providing a reproducible roadmap for anti-tumour immune engagement and immunotherapeutic stratification.
Open AccessDeciphering the Tumor Microenvironment: An Integrated Single-Cell RNA-Seq and AI Framework for Novel Biomarker and Therapeutic Target Discovery in Melanoma
Background: Melanoma represents a highly immunogenic and therapeutically challenging malignancy. The complex cellular ecosystem of the tumor microenvironment (TME) acts as a critical driver of immune evasion, patient prognosis, and treatment resistance. Traditional bulk sequencing often fails to resolve these high-resolution intercellular dynamics. This investigation presents an end-to-end explainable machine learning and network biology framework designed to deconstruct TME single-cell heterogeneity, discover novel candidate biomarkers, and map actionable cell-cell communication networks. Methodology: High-quality single-cell RNA sequencing (scRNA-seq) expression data from melanoma lesions (GSE115978) were processed using a multi-phase computational workflow. Following cellular filtration, library size normalisation, and highly variable gene (HVG) selection, cells were partitioned using unsupervised Leiden clustering and annotated via signature gene scoring matrix methods (Wolf et al., 2018; Traag et al., 2019). An optimal gradient-boosted tree ensemble classifier (XGBoost) was constructed for cross-compartment biomarker screening (Chen & Guestrin, 2016), interpreted using SHAP (SHapley Additive exPlanations) values (Lundberg & Lee, 2017), and augmented with a PyTorch deep autoencoder. Downstream systems analyses included biological pathway enrichment via gseapy (Kuleshov et al., 2016), ligand-receptor communication mapping via LIANA (Efremova et al., 2020; Dimitrov et al., 2022), and Pearson co-expression network profiling. Cross-cohort validation was conducted on an independent melanoma cohort (GSE72056) (Tirosh et al., 2016b), and prognostic utility was clinically validated using empirical patient survival data from the TCGA-SKCM cohort (TCGA Research Network, 2015; Davidson-Pilon, 2019). Results: Unsupervised Leiden clustering partitioned the cellular atlas into seven major structural and immunological compartments: Melanoma/Tumor, T-Cells, B-Cells, Macrophages, NK Cells, Endothelial cells, and Cancer-Associated Fibroblasts (CAFs). The trained XGBoost model prioritized CD79A, MLANA, LYZ, MFAP4 , and CDH5 as top candidate biomarkers across distinct microenvironmental niches. SHAP explainability analysis confirmed MLANA and S100B as key positive predictors of tumor cell identity, while highlighting a bidirectional distribution for B2M linked to antigen presentation downregulation. Functional enrichment mapped Coagulation and Epithelial-Mesenchymal Transition (EMT) as highly dysregulated pathways driving the malignant state. Cell-cell communication profiling inferred highly significant SERPINE1 signalling axes to LRP1 and PLAUR receptors. SPARC was identified as the central co-expression hub regulator. Cross-cohort screening in GSE72056 confirmed biomarker stability across independent patient profiles. Empirical clinical survival validation within the TCGA-SKCM cohort (n=314) demonstrated that heightened expression of CD79A correlates with a statistically significant survival advantage (Log-Rank p = 4.4202 × 10-3), extending median overall survival from 26.7 months to 44.8 months. Conclusion: This computational pipeline systematically resolves TME heterogeneity, revealing a robust biomarker signature centred on CD79A and MLANA, alongside SERPINE1-driven immune-stromal crosstalk. The discovery of the protective prognostic role of CD79A links single-cell immune networks directly to clinical patient outcomes, providing a reproducible roadmap for anti-tumour immune engagement and immunotherapeutic stratification.