Research Proposals

Multi-task prediction of kon, koff, and residence time via transfer learning from binding affinity.

Bridging sparse autoencoder interpretability and controllable generation in protein language models.

Integrating conformational ensembles into spatio-temporal graph learning for accurate protein–ligand scoring.

A comprehensive review, gap analysis, and proposed method for affinity prediction.

Interpretable feature steering for protein language model mutation effect prediction.

From genetic evolution and horizontal gene transfer to clinical manifestations and novel interventions.

De novo generation of diverse, potent, and pathogen-targeted AMPs against drug-resistant and fungal threats.

Morphological profiling in drug discovery with multi-scale self-supervised foundation models.

Integrating molecular graphs, multi-omics profiles, and biological knowledge for interpretable cancer therapy selection.

Integrating metagenomic profiles, host pharmacogenomics, and molecular structures for personalised pharmacotherapy.

Dynamic knowledge retrieval for grounded binding affinity estimation.

Cross-modal fusion of neuroimaging, genomics, and clinical data with calibrated uncertainty and missing-modality robustness.

Radiology report generation with selective prediction and confidence calibration.

High-fidelity, conditioned synthesis of multi-modal medical images to address data scarcity in rare diseases.

Privacy-preserving multi-institutional pre-training of vision transformers for cancer classification on WSIs.

Compact vision transformers for clinical-grade whole-slide image analysis.

DiffMRI-RC: rare-condition diffusion models for high-fidelity 3D brain MRI with anatomical and pathological control.

Masked autoencoder and contrastive pre-training on volumetric CT and MRI.

Mitigating catastrophic forgetting via sparse learnable priors and momentum-guided prompt updating in medical segmentation.

Test-time compute adaptation with Dirichlet uncertainty calibration for medical image classification.

A unified framework for multi-modal generative modelling with structural causal inference and perturbation prediction.

Diffusion models for trajectory inference from multi-omics time-series data.

Probabilistic chromatin accessibility inference with cell-type conditioning and regulatory interpretability.

Multimodal cross-attention fusion of DNA sequence and epigenomic profiles for enhancer, promoter, silencer, and insulator classification.

A scalable genomic foundation model with linear complexity and single-nucleotide resolution for variant effect prediction.

Multi-task deep learning for integrated 3D genome and epigenome profiling from Nanopore sequencing data.

Foundation model synthesis of spatial multi-omics from bulk RNA-seq for immune checkpoint blockade response prediction.

A diversity-aware adaptive framework for dynamic clinical environments and continual medical image classification.

Multi-evidence grounding for hallucination-resistant clinical question answering with large language models.

Bidirectional integration of deep learning and symbolic logic for explainable multi-modal diagnostic decision support.

Robust CRISPR-Cas9 off-target prediction with uncertainty-aware multi-view deep learning.

Precision CRISPR-Cas9 genome editing through unified gRNA efficiency and off-target prediction.

A metacognitive self-assessment framework for reliable uncertainty quantification and confidence calibration in large language models.

Iterative eligibility criteria refinement for clinical trials via real-world evidence simulation.

An explainable hybrid LSTM-Transformer with SHAP-LIME integration for interpretable market forecasting.