http://repository.iiitd.edu.in/xmlui/handle/123456789/1243 CB Sun, 30 Aug 2026 14:35:49 GMT 2026-08-30T14:35:49Z http://repository.iiitd.edu.in/xmlui/handle/123456789/2039 FMT hub phase II: comprehensive web infrastructure for the FMT center and the microbiome informatics lab at IIIT-Delhi Bhaskar, Siddharth; Bali, Siddhant; Kumar, Rishabh; Ghosh, Tarini Shankar (Advisor) We present FMT Hub - Phase II, We are developing the official website for the Microbiome Informatics Lab at IIIT Delhi as a digital extension of the FMT Hub. The platform is being built uscentralized 15, React 19, Node.js, Tailwind CSS, and TypeScript, focusing on modular UI, SEO, and responsiveness. We’re implementing secure APIs, dynamic routing, and scalable backend architecture for seamless performance. Deployment is handled on Oracle Linux with Apache, using SSL-secured HTTPS. This site will serve as a centralized digital gateway for lab research, data resources, and collaborations. Thu, 17 Jul 2025 00:00:00 GMT http://repository.iiitd.edu.in/xmlui/handle/123456789/2039 2025-07-17T00:00:00Z http://repository.iiitd.edu.in/xmlui/handle/123456789/2038 Unsupervised clustering of kinases based on sequence and structural features Yadav, Karan; Singh, Manveet; Gautam, Mayank; Ray, Arjun (Advisor) The way kinases interact with their substrates has been a longstanding problem in biology prediction and important to drug developing. This report is about a new computational system like Multimodal Siamese Graph Network that is helpful in predicting the ability of kinase substrate engagement effectively, using different data modalities. Our model is on both sequence and structural based features. The structural information in PDB files has been mapped into a graphical form and analyzed by structural features (binding pockets and secondary structures) along with sequence information as pre-processed structural information (including ESM-2 protein language model embeddings). Siamese network architecture is learning to differentiate among interacting pairs, and non-interacting pairs using a triplet Margin Loss syntax. On a held out validation set our model achieves a state of the art Area Under the Precision-Recall Curve (AUPRC) and Area Under the Receiver Operating Characteristic (AUROC) of 0.785 and 0.786, respectively, far outpacing the baseline models, including Random Forest and XGBoost. Impressively, the model is unsupervised in that the kinases in their existing biological families are clustered which implies that the network is learning biologically important representations. The application proves not only the effectiveness of the combination of deep sequence, structural information to model molecular recognition but also represents a solid foundation towards the future in silico discoveries of the phosphorylation phenomena by means of generative models Thu, 24 Jul 2025 00:00:00 GMT http://repository.iiitd.edu.in/xmlui/handle/123456789/2038 2025-07-24T00:00:00Z http://repository.iiitd.edu.in/xmlui/handle/123456789/2037 Allergen based classification of peptides Jindal, Naman; Mehta, Neil; Saini, Saksham; Bagler, Ganesh (Advisor) Peptides, as essential biological molecules, exhibit diverse structural and functional properties, some of which can trigger allergenic responses and pose significant challenges in biotherapeutic applications and drug design. This project investigates the binary classification of peptides into liable (allergenic) and non-liable (non-allergenic) categories using advanced machine learning techniques, aiming to aid in early allergenicity screening during drug development and bioengineering applications. The research utilized a comprehensive dataset of 25,000 peptide sequences to develop an end-to-end machine learning pipeline for allergenic potential prediction. The methodology involved extensive feature extraction using multiple biochemical and structural descriptors including Amino Acid Composition (AAC), Dipeptide Composition (DPC), Tripeptide Composition (TPC), Composition of k-Spaced Amino Acid Pairs (CKSAAP) and various grouped compositions computed through the iFeature tool. Additionally, ESM2 transformer-based language model embeddings were incorporated to capture deep contextual patterns. Many feature selection strategy was implemented, combining filter methods (Variance Threshold and ANOVA F-score), embedded methods (XGBoost) in a multi-stage process to optimize the feature space while preserving critical information. The framework was evaluated using a comprehensive suite of machine learning and deep learning models, including Random Forest, Support Vector Machines (SVM), K-Nearest Neighbors (KNN), Gradient Boosting variants (CatBoost, AdaBoost, XGBoost, LightGBM), and deep learning architectures such as Multilayer Perceptron (MLP), Convolutional Neural Networks (CNN), Recurrent Neural Networks (RNN), and ResNet. Model performance was rigorously assessed using multiple evaluation metrics including Accuracy, AUROC, AUPRC, F1-score, Cohen's Kappa, and Matthews Correlation Coefficient (MCC). This project demonstrates the potential of combining advanced bioinformatics techniques with machine learning to address critical biological classification challenges and establishes a foundation for developing practical tools for peptide allergenicity screening in therapeutic and food applications. Fri, 18 Jul 2025 00:00:00 GMT http://repository.iiitd.edu.in/xmlui/handle/123456789/2037 2025-07-18T00:00:00Z http://repository.iiitd.edu.in/xmlui/handle/123456789/2036 Fine-tuning language models for novel recipe generation Gautam, Sambhav; Singh, Sambhav; Bagler, Ganesh (Advisor) This project focuses on fine-tuning four lightweight language models—Phi-1, Phi-1.5, Phi-2, and TinyL- lama—for novel recipe generation using a recipe dataset sourced from MongoDB. The process involved data preprocessing, tokenization, fine-tuning with Parameter-Efficient Fine-Tuning (PEFT) and Quan- tized Low-Rank Adaptation (QLoRA), and evaluation using BLEU, METEOR, BERTScore, and Per- plexity metrics. The results show that Phi-1.5 achieved the best performance (BLEU: 0.7399, METEOR: 0.9281, BERTScore: 0.9533, Perplexity: 39.55), indicating high-quality and coherent recipe generation. TinyLlama exhibited potential hallucination due to its high perplexity (5044.15). This work demon- strates the efficacy of efficient fine-tuning techniques for domain-specific text generation in culinary applications. Tue, 01 Jul 2025 00:00:00 GMT http://repository.iiitd.edu.in/xmlui/handle/123456789/2036 2025-07-01T00:00:00Z