End-to-end web platform for comparative structural analysis of protein–protein complexes, supporting interface characterization, energetic analysis, and variant-level comparison.

Ab-Affinity, a language model that can accurately predict the binding affinity of antibodies against a target peptide, e.g., the SARS-CoV-2 spike protein. The model achieved a peak Pearson correlation coefficient of 0.712 and successfully mapped therapeutic antibody thermostability without explicit training.

Structure-aware ESM2 framework that jointly predicts antibody binding and neutralization for SARS-CoV-2 variants. Improves binding AUC 0.84 → 0.93 and neutralization AUC 0.88 → 0.95, outperforming DeepAIR, AntiBERTa, AbMap, and A2Binder.

LLM-based framework trained on antigen-BCR sequence pairs. Accurately predicts antigen-specific B-cell receptors from full repertoires, generalizes to unseen antigens, and validates candidates via molecular dynamics simulations.

Ab-Affinity: hybrid LLM-driven antibody design framework integrating genetic algorithms and simulated annealing to generate novel SARS-CoV-2-targeting antibodies. Achieves >160× binding affinity improvements over experimental candidates while maintaining strong biophysical stability.

Lightweight breast cancer histopathology classifier combining self-supervised contrastive learning and a ResNet–Inception architecture. Achieves up to 98% accuracy with ~50% fewer parameters than existing leading models.

Benchmarked 27 classifiers for SARS-CoV-2 3CLpro inhibitor prediction. Neural network achieves 91–93% accuracy. SHAP-based interpretability identifies key molecular fingerprints for antiviral drug screening. Validated on >100,000 molecules.

Data-efficient deep learning architecture for yoga pose recognition capturing spatial and depth features. Outperforms ResNet, Inception, and Xception; achieves 94.9% accuracy and 95.6% precision.

Deep learning model for Bangla-language fake news classification, addressing low-resource NLP challenges in misinformation detection.

Meta-heuristic-enhanced K-means clustering that improves convergence and cluster quality for high-dimensional, multi-cluster datasets.