http://repository.iiitd.edu.in/xmlui/handle/123456789/1793 Year-2026 2026-08-30T16:16:04Z http://repository.iiitd.edu.in/xmlui/handle/123456789/2030 Development of AI-based tools for functional annotation of lncRNA and its applications Choudhury, Shubham; Raghava, Gajendra Pal Singh (Advisor) Long non-coding RNAs (lncRNAs) are a significant component of the human transcriptome and are well known for their ability to regulate gene expression. They have been implicated in a variety of disease-associated pathways. However, functional annotation of these lncRNAs remains a major challenge in the era of genomics and metagenomics, where large-scale genome sequencing has become routine. In this thesis, we have attempted to develop computational resources for annotating lncRNAs and to explore their applications in cancer diagnostics. In the first part of this thesis, we compiled all lncRNAs resources that includes experimental techniques, prediction methods and data repositories. These resources were manually curated from the literature and made available to the scientific community through a web platform called lncInfo. Next, we developed CytoLNCpred, a cell line–specific computational method for identifying cytoplasm-associated lncRNAs across 15 human cell lines. We implemented multiple computational approaches, including machine learning, deep learning, and large language models; in order to achieve improved prediction accuracy. The second part of the thesis describes a method, lncrnaPI, developed to predict lncRNA–protein interacting pairs using transformer-based sequence embeddings as features. In the final section of the thesis, we explored the cancer diagnostic potential of lncRNAs based on their transcriptomic profiles. Specifically, we analyzed lncRNA transcriptomic profiles in patients with pancreatic ductal adenocarcinoma (PDAC). We explored two independent approaches for identifying lncRNA-based biomarkers capable of discriminating PDAC patients from healthy individuals. First, we developed machine learning–based models using traditional approaches for biomarker identification. Second, we investigated a new concept in which large language models were used for mining transcriptomic data. One key objective of this research work is to facilitate the scientific community working in the field of lncRNA biology. Therefore, we developed web-based servers and standalone tools that are publicly available fostering open science. The work presented in this thesis provides novel computational strategies for enhancing the functional annotation of lncRNAs, decoding their molecular interactions, and exploiting their expression patterns for clinically meaningful applications. 2026-07-01T00:00:00Z http://repository.iiitd.edu.in/xmlui/handle/123456789/2029 Development of AI-based tools for designing subunit vaccines Kumar, Nishant; Raghava, Gajendra Pal Singh (Advisor) Immunization continues to be a cornerstone of global public health, drastically reducing mortality rates worldwide. Over time, the field of vaccinology has progressively shifted from utilizing whole-pathogen preparations toward highly targeted epitope and subunit formulations. However, traditional laboratory-based methods for vaccine discovery remain heavily constrained by high financial costs and extended developmental timelines. To mitigate these bottlenecks, this thesis introduces a comprehensive suite of computational resources for designing epitope or subunit based vaccines. The work presented in this thesis is organized into three major components: identification of epitope-based vaccine candidates, safety evaluation of predicted candidates, and assessment of stability under physiological and environmental conditions. Under the first objective, a tool named CLBtope was developed to predict both linear and conformational B-cell epitopes from antigen sequences, facilitating the identification of potential vaccine candidates capable of eliciting protective B-cell and antibody responses. To ensure candidate safety, two predictive models were established: DMPPred, which identifies peptides associated with type 1 diabetes mellitus, and AlgPred3, which comprehensively evaluates the allergenic risks of proposed vaccine components. Finally, to address peptide and protein stability under diverse physiological and environmental conditions, three distinct platforms were developed: AFProPred for the characterization and design of antifreeze proteins, PPTStab for engineering thermostable proteins, and PlifePred2 for estimating peptide half-lives in body fluids. All tools developed in this thesis are freely available to the scientific community as web servers and standalone software through platforms such as GitHub and PyPI. State-of-the-art AI methodologies, including machine learning, deep learning, and large language models, were deployed to design effective subunit vaccines. 2026-03-01T00:00:00Z http://repository.iiitd.edu.in/xmlui/handle/123456789/1995 SAFE-ICU: advancing AI-driven decision support in the intensive care units Singh, Pradeep; Sethi, Tavpritesh (Advisor); Lodha, Rakesh (Advisor) Integrating artificial intelligence into healthcare presents an unprecedented opportunity to address critical challenges in Intensive Care Units. The objective of this thesis is to create, validate, and deploy AI models to increase safety through the development of early warning systems for drug-drug interactions and predicting clinical decompensation of patients. Our first objective is creation of a data lake involving multimodal data, which combines vital signs, laboratory tests, treatment histories, patient files, and thermographic videos; not just images. The combination of structured and unstructured data contained in the datasets offers unmatched insight into intensive care unit patients, thereby enhancing the overall quality of research. The datasets used for prediction and diagnostics are pre-processed through various techniques like feature extraction, temporal alignment, etc. The second Objective aims at the creation of actionable, predictive, and management tools for sophisticated critical care of ICU patients. For example, ThermalShockNet, which was created for pediatric ICU settings and utilizes thermal videos and patient vitals to identify subtle physiological changes indicative of hemodynamic shock. This model uses data from thermal frames and vital signs to make predictions. When combined using deep learning, these predictions can reveal the onset of shock before it appears in a patient. ThermoGnose, a pipeline developed specifically for the early prediction of hypothermia, uses non-linear features drawn from patient monitoring data to estimate hypothermia onset and provide vital lead time for clinical intervention. SIgnose predicts shock risk by analyzing vital sign trends. When combined, this variety of approaches highlight how AI can address complex challenges in the ICU. The third Objective was to improve drug safety in the ICU. For this, the ContraIndicator tool was created. NLP processes the treatment chart and line listing of drugs to find probable adverse drug–drug interactions. The ContraIndicator insights assist clinicians in minimizing the chance of adverse reactions. The end Objective was SafeICU, a modular and explainable AI-driven real-time platform for the pediatric ICU. It brings together early warning systems along with safety and infection surveillance modules, which include vital signs, thermal imaging, medication, and microbiology results. At SafeICU, they help flag unusual Shock Index values in many ICUs. The ContraIndicator is on the lookout for drug interactions through the examination of ICU prescriptions. AMRtrack monitors the emergence of antimicrobial resistance by processing laboratory reports, mapping trend data, and informing infection control planning. When these modules are together, real-time alerts, risk scores, and infection monitoring are delivered into the ICU workflow. Overall, this thesis provides a clinically oriented framework for using AI in ICUs. It shows how models that can be multimodal, explainable, and deployable work together in real time to assist clinical decision making in high-acuity clinical settings. These advances create a pathway to the secure and scalable exploitation of AI in critical care and, eventually, precision medicine in health care that are rich in data. 2026-05-01T00:00:00Z http://repository.iiitd.edu.in/xmlui/handle/123456789/1964 Mechanism-informed, AI-driven frameworks for discovery and validation of aging-associated chemical space Arora, Sakshi; Ahuja, Gaurav (Advisor) Aging is a progressive, multifactorial biological process that drives the risk of nearly all major chronic diseases, including cancer, neurodegeneration, metabolic disorders, frailty, and cardiovascular dysfunction. Although the past two decades have established a molecular framework through the Hallmarks of Aging, translating this knowledge into actionable, small-molecule interventions that enhance healthspan remains a central challenge in the field of geroscience. Experimental discovery pipelines are slow, resource-intensive, and typically explore only a minute fraction of chemical space. Conversely, computational drug discovery approaches, while high-throughput, often rely on chemistry-centric descriptors, exhibit black-box behavior, lack mechanistic interpretability, and rarely generalize to biologically novel molecules. This thesis addresses these long-standing limitations by developing two complementary artificial intelligence systems, AgeXtend and AgeXtend:: Mimetics, designed to accelerate mechanism-informed discovery of geroprotective molecules and caloric restriction mimetics (CRMs). The first objective introduces AgeXtend, a multimodal, bioactivity-driven, and fully explainable AI framework. AgeXtend integrates curated datasets of experimentally validated geroprotectors and neutral compounds with bioactivity-based descriptors, hallmark-specific classification models, toxicity prediction, and target inference modules. By combining mechanistic knowledge with machine learning, AgeXtend achieves robust predictive accuracy across cross-validation, leave-one-out validation, and independent external datasets. Importantly, the explainability module maps predictions onto nine aging pathways, allowing for a mechanistic interpretation of each compound’s mode of action. Large-scale screening of ~1.1 billion compounds yielded diverse chemical classes with strong geroprotective potential. Experimental validation confirmed these predictions across three biological systems: Saccharomyces cerevisiae chronological lifespan assays, human fibroblast senescence assays, and Celegans lifespan assays. Endogenous metabolites and repurposed drugs predicted by AgeXtend demonstrated lifespan-extending or senomodulatory activity, underscoring the biological fidelity of its predictions. Building upon this foundation, the second objective presents AgeXtend::Mimetics, a novel computational framework designed to identify Caloric Restriction Mimetics, compounds capable of reproducing CR-like physiological responses without structural similarity to known CRMs. Unlike existing approaches that rely on transcriptomic signatures alone or structural matching, AgeXtend::Mimetics explicitly decouples biological convergence from chemical divergence. Using dual similarity modeling, ridge regression residuals, supervised contrastive learning, and composite CRM fingerprinting, the framework identifies molecules whose biological signatures align strongly with known CRMs despite having distinct chemical architectures. Large-scale application across thousands of compounds revealed chemically novel, mechanistically plausible CRM candidates that align with pathways such as nutrient sensing, autophagy, mitochondrial remodeling, and metabolic regulation. This framework substantially broadens the chemical landscape of CRM discovery and provides mechanistic clarity on CRM-like effects. Together, the approaches developed in this thesis demonstrate that explainable, mechanism-oriented AI models can successfully bridge the gap between large-scale chemical exploration and biological relevance. AgeXtend and AgeXtend::Mimetics collectively advance the field of computational geroscience by enabling scalable, interpretable, and experimentally validated discovery of geroprotectors and CRMs. These contributions lay the groundwork for future translational studies, the development of generative design for longevity therapeutics, and the integration of multi-omic datasets to refine mechanism-based discovery pipelines. The thesis highlights both the promise and current limitations of AI in aging biology, providing a roadmap for next-generation computational frameworks that target healthspan extension. 2026-03-01T00:00:00Z