Subhasankar Khilar
Research Subjects
Research Interests
Subhasankar Khilar has a total of 1 co-authors. Here are the top co-authors by number of shared publications:
- Elamathi Natarajan
Bioinformatics CRO Scientist, Biotecnika, Bangalore, India
2 shared publications
Latest Publications
- 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
Graph Neural Networks (GNNs) for Protein-Ligand Interaction Prediction
Predicting protein-ligand interactions in the modern drug discovery has revolved from the involvement of artificial intelligence and structural bioinformatics using Graph Neural Networks (GNNs). The limited explainability of GNN models presents an important encumbrance in biomedical research, but it has achieved a high degree of accuracy in determining and identifying binding affinity and active compounds, as evidenced by [1] [2] [3] [4]. Here this research focuses on the interpretation of protein-ligand interactions at a molecular level, a rapidly developing area within Graph Neural Networks (GNNs). Now days modern study handling techniques such as visualization techniques, attention mechanism and model-based feature ascription by model to boost, and make robust and decrease false predictions on binding. Along with some approaches include like graph pooling strategies, message-passing optimization, self-supervised learning, transfer learning and contrastive learning are rapidly utilized to enhance the representative learnings. Furthermore, integration of molecular docking simulations, hybrid deep learning architectures and protein language model gives more reliable & biological predictions of protein-ligand interactions. That focuses on given process that identifies key ligand atoms and binding residues, as well as physicochemical factors influencing affinity, through chemical thought processes. Here this research work identified the challenges of developing biologically significant explanations, transparency, and the corollary dataset biases on interpretability. The research work conducted an in-depth investigation into the consolidation of protein language models to establish more reliable pathways for future research, examining hybrid architectures, transparent and energy-efficient GNNs, and scientifically grounded AI models for drug discovery. My research work highlights that XGNNs establishes a connection between Deep Learning and Biochemical expertise with increased confidence, which will enhance the accuracy of predictive models and computational models.
Open AccessGraph Neural Networks (GNNs) for Protein-Ligand Interaction Prediction
Predicting protein-ligand interactions in the modern drug discovery has revolved from the involvement of artificial intelligence and structural bioinformatics using Graph Neural Networks (GNNs). The limited explainability of GNN models presents an important encumbrance in biomedical research, but it has achieved a high degree of accuracy in determining and identifying binding affinity and active compounds, as evidenced by [1] [2] [3] [4]. Here this research focuses on the interpretation of protein-ligand interactions at a molecular level, a rapidly developing area within Graph Neural Networks (GNNs). Now days modern study handling techniques such as visualization techniques, attention mechanism and model-based feature ascription by model to boost, and make robust and decrease false predictions on binding. Along with some approaches include like graph pooling strategies, message-passing optimization, self-supervised learning, transfer learning and contrastive learning are rapidly utilized to enhance the representative learnings. Furthermore, integration of molecular docking simulations, hybrid deep learning architectures and protein language model gives more reliable & biological predictions of protein-ligand interactions. That focuses on given process that identifies key ligand atoms and binding residues, as well as physicochemical factors influencing affinity, through chemical thought processes. Here this research work identified the challenges of developing biologically significant explanations, transparency, and the corollary dataset biases on interpretability. The research work conducted an in-depth investigation into the consolidation of protein language models to establish more reliable pathways for future research, examining hybrid architectures, transparent and energy-efficient GNNs, and scientifically grounded AI models for drug discovery. My research work highlights that XGNNs establishes a connection between Deep Learning and Biochemical expertise with increased confidence, which will enhance the accuracy of predictive models and computational models.