Predicting C- and S-linked Glycosylation Sites from Protein Sequences Using Protein Language Models
(Published)
Developed a hybrid deep-learning architecture to predict C- and S-linked glycosylation sites from protein sequences using protein language model embeddings and contextual information. Link to article: ScienceDirect

StackGlyEmbed: Prediction of N-linked Glycosylation Sites Using Protein Language Models
(Published)
Proposed StackGlyEmbed, a model that predicts N-linked glycosylation sites from protein sequences by leveraging protein language models with window and per-residue features. Link to article: Bioinformatics Advances

ResLysEmbed: A ResNet-Based Framework for Succinylated Lysine Residue Prediction Using Sequence and Language Model Embeddings
(Published)
Developed a hybrid deep-learning architecture incorporating protein language models to identify succinylated lysine residues. Link to article: Bioinformatics Advances

NFEmbed: Modeling Nitrogenase Activity via Classification and Regression with Pretrained Protein Embeddings
(Published)
Developed stacking ensemble models to predict microbial strains with high nitrogenase potential using protein language model embeddings and achieved superior performance over state-of-the-art methods. Link to article: Bioinformatics Advances

Predicting RNA 5-Hydroxymethylcytosine Modification with Deep Learning Models Using RNA Language Model Embeddings
(Under Revision)
Designed a dual-branch deep learning model architecture to predict RNA 5-Hydroxymethylcytosine modifications using RNA language models and extracted biological interpretations.

DeepBCTPred: Deep Learning-Based Prediction of Bladder Cancer Tissues from Endoscopic Images
(Under Review — CSE472 Machine Learning Project)
Developed a pipeline to generate new images and a novel genetic algorithm to effectively select images from them and combined handcrafted features with learned features from convolutional neural networks.

Prediction of Protein-Carbohydrate Binding Sites from Protein Primary Sequence
(Under Rebuttal)
Developed StackCBEmbed, an ensemble machine learning model for effective classification of protein-carbohydrate binding interactions at the residue level, and integrated sequence-based features with pre-trained transformer-based protein language model embeddings. Preprint: bioRxiv

Predicting Protein-Carbohydrate Binding Sites: A Deep Learning Approach Integrating Protein Language Model Embeddings and Structural Features
(Under Review — Undergraduate Thesis)
Designed a novel deep-learning architecture that integrates protein language model embeddings with structural features for predicting protein-carbohydrate binding sites.

Expanded Strategy Space Improves Nash Solution by Increased Degrees of Freedom
(Manuscript in Preparation — CSE462 Algorithm Engineering Project)
Investigated algorithmic improvements for solving the Nash Equilibrium problem and focused on approximation algorithms and meta-heuristic approaches, such as replicator dynamics, to enhance computational efficiency.

OptEmbed: A Multi-Task Framework Using Protein Language Model Embeddings and Sequential Features to Predict Optimal Temperature, Melting Temperature, and Optimal pH of Enzymes
(Manuscript in Preparation)
Introduced OptEmbed, a protein language model–based framework integrating sequential features to predict enzyme parameters (Topt, Tm, pHopt) with improved accuracy and interpretability over state-of-the-art methods.

BUET RISE Grant
RISE Student Research Grant [No. S2024-01-004] (Grant received)

HONORABLE MENTIONS
MicroProcessor and MicroController project named Unbeatable Protection: The 5-way Security Vault