Ai Bioinformatics

Deep Learning Protein Structure Prediction

Development of neural network architectures for accurate prediction of three-dimensional protein folding from amino acid sequences.

RESEARCH GAP FRONTIERS

Transformer Models for Genomic Sequence Analysis

Application of transformer-based architectures to identify regulatory elements and predict gene expression from DNA sequences.

RESEARCH GAP FRONTIERS

Graph Neural Networks in Molecular Docking

Utilization of graph-based deep learning for predicting protein-ligand binding affinity and spatial interactions.

RESEARCH GAP FRONTIERS

Attention Mechanisms for Drug Discovery

Implementation of attention-based models to identify key molecular features driving drug efficacy and toxicity.

RESEARCH GAP FRONTIERS

Reinforcement Learning for Protein Engineering

Use of reinforcement learning algorithms to optimize protein sequences for enhanced functional properties and stability.

RESEARCH GAP FRONTIERS

Generative Models for De Novo Drug Design

Development of variational autoencoders and GANs to generate novel drug candidates with desired pharmacological properties.

RESEARCH GAP FRONTIERS

Federated Learning for Genomic Privacy

Distributed machine learning approaches enabling collaborative genomic analysis while maintaining patient data privacy and security.

RESEARCH GAP FRONTIERS

Multi-modal AI Integration Biomedical Data

Fusion of diverse omics data types through multi-modal neural networks for comprehensive disease understanding.

Interpretable Machine Learning in Clinical Genomics

Development of explainable AI models that identify clinically actionable genetic variants and provide interpretable predictions.

Causal Inference in Gene Regulatory Networks

Application of causal inference methods to infer true regulatory relationships and gene-to-gene dependencies from omics data.

Graph Convolutional Networks for Protein Interactions

Modeling protein-protein interaction networks using graph convolutional architectures to predict functional associations.

Transfer Learning Across Species Genomics

Leveraging pre-trained models across evolutionary distant species to improve prediction accuracy with limited labeled data.

Zero-shot Learning for Rare Disease Diagnosis

Development of AI models capable of diagnosing unseen rare diseases by transferring knowledge from known genetic disorders.

Adversarial Robustness in Biomedical Predictions

Investigation of vulnerabilities and development of robust AI models resilient to adversarial inputs in clinical applications.

Uncertainty Quantification in Genomic Analysis

Integration of Bayesian methods and probabilistic modeling to quantify confidence intervals in variant calling and predictions.

Temporal Deep Learning for Disease Progression

Application of recurrent neural networks and temporal convolutional networks to model longitudinal patient omics trajectories.

Few-shot Learning for Personalized Medicine

Meta-learning approaches enabling rapid adaptation to individual patient genomes with minimal training examples.

Knowledge Graph Embedding for Drug Repurposing

Construction and embedding of biomedical knowledge graphs to identify novel therapeutic applications for existing drugs.

Attention-based RNA Secondary Structure Prediction

Development of attention mechanisms to capture long-range dependencies in RNA sequences for structure prediction.

Contrastive Learning in Protein Space

Self-supervised learning approaches to learn protein representations by maximizing similarity between functionally related proteins.

Equivariant Neural Networks for Structural Biology

Development of rotation and translation equivariant architectures for predicting protein structures and molecular properties.

Quantum Machine Learning for Drug Interactions

Exploration of quantum computing paradigms to model complex quantum mechanical effects in molecular interactions.

Meta-learning for Variant Effect Prediction

Application of meta-learning algorithms to predict functional consequences of novel genetic variants with limited homologous examples.

Graph Attention Networks for Pathway Analysis

Implementation of graph attention mechanisms to identify key genes and regulatory interactions in biological pathways.

Neural Ordinary Differential Equations Biological Dynamics

Utilization of neural ODEs to model continuous biological processes and predict cellular state transitions.

Diffusion Models for Molecular Generation

Application of diffusion-based generative models to design novel bioactive molecules with specified properties.

Active Learning for Annotation Efficiency

Strategic selection of uncertain samples for annotation to maximize model performance with minimal experimental validation.

Anomaly Detection in Genomic Sequencing

Development of unsupervised and semi-supervised methods to identify sequencing artifacts and biological outliers.

Self-supervised Learning for Unlabeled Omics

Pretraining strategies on large unlabeled biomedical datasets to improve downstream task performance.

Capsule Networks for Hierarchical Protein Features

Implementation of capsule networks to capture hierarchical relationships between protein domains and functional modules.

Bayesian Deep Learning for Clinical Predictions

Integration of Bayesian inference with deep learning to provide uncertainty estimates for patient risk stratification.

Explainable AI for Mutation Functional Impact

Development of interpretable models that explain how genetic mutations affect protein function through molecular mechanisms.

Time-series Forecasting for Protein Evolution

Application of temporal modeling to predict evolutionary trajectories and adaptive mutations in protein families.

Mixture of Experts for Multi-task Genomics

Development of adaptive ensemble models that specialize in different genomic prediction tasks simultaneously.

Hypergraph Learning for Complex Biological Networks

Application of hypergraph neural networks to model higher-order interactions in biological systems beyond pairwise relationships.

Domain Adaptation for Cross-cohort Genomics

Development of techniques to transfer models trained on one population to accurately predict phenotypes in genetically distinct populations.

Sparse Learning for Feature Selection Omics

Utilization of sparsity-inducing methods to identify minimal sets of biomarkers with maximum predictive power.

Topological Data Analysis for Molecular Complexity

Application of persistent homology and topological methods to uncover hidden structures in high-dimensional molecular data.

Vision Transformers for Microscopy Image Analysis

Implementation of transformer-based architectures for analyzing cellular and tissue images with high interpretability.

Natural Language Processing for Biomedical Literature

Application of NLP techniques to extract and integrate knowledge from biomedical literature for hypothesis generation.

Symbolic Reasoning Integration with Deep Learning

Combination of logical reasoning with neural networks to enforce biological constraints in predictions.

Attention-based Multiple Instance Learning Pathology

Development of weakly-supervised models using attention mechanisms to identify diagnostic regions in whole-slide images.

Curriculum Learning for Complex Biological Tasks

Implementation of progressive training strategies that gradually increase task complexity to improve model generalization.

Ensemble Methods for Robust Biomarker Discovery

Application of ensemble learning approaches to identify stable and reproducible biomarkers across multiple datasets.

Graph Pooling for Molecular Property Prediction

Development of hierarchical graph pooling mechanisms to predict molecular properties while maintaining interpretability.

Continual Learning for Evolving Genomic Databases

Design of models that continuously adapt to new genomic data without catastrophic forgetting of prior knowledge.

Fairness and Bias Mitigation in Clinical AI

Development of debiasing techniques to ensure equitable performance of AI models across diverse patient populations.

Neural Implicit Representations for Protein Dynamics

Use of neural implicit functions to model continuous conformational spaces and protein dynamics trajectories.

Metric Learning for Genomic Similarity Search

Development of learned distance metrics for efficient similarity search and clustering of genomic sequences.

Structural Alignment Learning for Conserved Elements

Training of neural networks to learn meaningful structural alignments for identifying functionally conserved genomic regions.

Recurrent Neural Networks Time-Series Omics

Development of RNN architectures for temporal analysis of longitudinal multi-omics data in disease progression and treatment response.

Variational Autoencoders Single-Cell Biology

Application of VAE frameworks to model cellular heterogeneity and latent biological factors in single-cell RNA-sequencing datasets.

Protein Language Models Functional Annotation

Leveraging large-scale protein language models pre-trained on sequence databases for zero-shot functional annotation of novel proteins.

Graph Isomorphism Networks Metabolite Prediction

Employing graph isomorphism neural networks to predict metabolite transformations and pathway intermediates in biochemical reactions.

Attention-based Chromatin Interaction Modeling

Development of attention mechanisms to capture long-range chromatin contacts and 3D genome organization from Hi-C data.

Integrative Deep Learning Multi-Omics Fusion

Design of neural architectures that jointly integrate genomics, proteomics, metabolomics, and imaging data for systems-level biological understanding.

Adversarial Learning Synthetic Biological Data Generation

Application of GANs to generate realistic synthetic genomic and proteomic datasets while preserving biological constraints and statistics.

Deep Metric Learning Antibody Affinity Prediction

Implementation of metric learning approaches to predict antibody-antigen binding affinities by learning meaningful biological distance functions.

Reinforced Learning Drug Combination Optimization

Application of RL agents to discover optimal multi-drug combinations that minimize toxicity while maximizing therapeutic efficacy.

Sequence-to-Sequence Models CRISPR Design

Utilization of encoder-decoder architectures to predict optimal CRISPR guide RNA sequences and assess off-target binding potential.

Hierarchical Bayesian Models Disease Subtyping

Development of hierarchical Bayesian frameworks to identify clinically relevant disease subtypes from high-dimensional patient omics data.

Point Cloud Neural Networks Protein Surface Analysis

Application of point cloud deep learning methods to analyze protein surface properties and predict ligand binding sites.

Normalizing Flows Posterior Inference Genetics

Employing normalizing flow models for efficient posterior inference in population genetics and phylogenetic inference problems.

Attention Pooling Genomic Region Classification

Design of attention-based pooling mechanisms to aggregate signals from multiple genomic regions for enhancer and promoter classification.

Neural Architecture Search Biomarker Discovery

Application of NAS methods to automatically design optimal neural architectures for identifying disease-specific biomarkers from omics data.

Contrastive Divergence Learning Sequence Motifs

Implementation of contrastive learning approaches to discover functional sequence motifs and regulatory elements in genomic data.

Probabilistic Graphical Models Gene Regulation

Development of probabilistic graphical models including Bayesian networks and Markov random fields for inferring gene regulatory relationships.

Representation Learning Cellular Images Phenotypes

Development of self-supervised representation learning methods to extract meaningful cellular phenotypes from high-resolution microscopy images.

Manifold Learning Genomic Data Visualization

Application of advanced manifold learning techniques beyond t-SNE and UMAP for interpretable visualization of complex genomic datasets.

Molecular Fingerprinting Deep Learning Compounds

Development of learnable deep molecular fingerprint representations for improved molecular property prediction and compound clustering.

Entropy-based Feature Selection Omics Analysis

Application of information-theoretic feature selection methods to identify non-redundant and biologically meaningful features from high-dimensional omics.

Attention Mechanisms Epitope Prediction Immunology

Design of attention-based neural models to predict immunogenic epitopes and T-cell receptor binding specificity from sequence data.

Subgraph Neural Networks Pathway Extraction

Application of subgraph neural networks to automatically extract and validate biochemical pathways from protein interaction networks.

Neural ODE Personalized Disease Progression Models

Implementation of continuous-time neural differential equations to model personalized patient-specific disease trajectories and prognosis.

Epistasis Detection Deep Learning Networks

Development of deep learning architectures to detect complex gene-gene interaction effects and epistatic relationships from genomic data.

Sequence Embedding Integration Phenotype Prediction

Creation of unified embedding spaces combining sequence and phenotypic information for improved genotype-phenotype association prediction.

Variational Inference Population Genomics Inference

Application of variational inference techniques for scalable Bayesian inference in large-scale population genomics datasets.

Multi-task Learning Disease Risk Stratification

Development of multi-task learning frameworks to jointly predict multiple disease outcomes and risk factors from clinical data.

Causal Discovery Molecular Interaction Networks

Application of causal inference algorithms to infer directional relationships and causal mechanisms in molecular interaction networks.

Attention Visualization Bioinformatics Model Interpretation

Development of visualization techniques to interpret attention patterns in deep bioinformatics models and identify important biological features.

Semi-supervised Learning Disease Classification

Application of semi-supervised learning to leverage both labeled and unlabeled patient data for improved disease classification and diagnosis.

Protein Interaction Prediction Graph Embeddings

Development of graph embedding methods to predict novel protein-protein interactions by learning latent representations from interaction networks.

Ensemble Clustering Genomic Data Stratification

Application of ensemble clustering approaches to robustly identify stable and biologically meaningful patient subgroups from omics data.

Tensor Factorization Drug-Target Interactions

Implementation of tensor decomposition methods to predict novel drug-target interactions by learning from multi-modal biological data.

Active Learning Rare Variant Classification

Application of active learning strategies to efficiently prioritize annotation of rare genetic variants with limited training data.

Equivariance Symmetry Molecular Property Learning

Development of neural networks that respect molecular symmetries and equivariance principles for improved molecular property prediction.

Survival Analysis Deep Learning Clinical Outcome

Development of deep survival analysis models that predict patient survival curves and time-to-event outcomes from clinical and omics data.

Graph Regression Networks Binding Energy Prediction

Implementation of graph regression models to predict molecular binding energies and binding affinities from molecular structure representations.

Clustering Quality Assessment Genomic Subgroups

Development of metrics and validation frameworks to assess biological validity and clinical relevance of computational genomic clusters.

Natural Language Processing Clinical Notes Extraction

Application of NLP methods to extract structured phenotypic and clinical information from unstructured electronic health records text.

Matrix Factorization Gene Expression Prediction

Implementation of advanced matrix factorization techniques to predict missing gene expression values and impute sparse omics matrices.

Spectral Methods Biological Network Analysis

Application of spectral clustering and spectral graph methods to decompose complex biological networks into functional modules.

Cross-modal Retrieval Biomedical Images Text

Development of cross-modal learning models linking histopathology images with clinical reports for improved diagnostic support.

Optimal Transport Omics Data Alignment

Application of optimal transport theory to align and integrate multi-batch omics datasets while preserving biological variation.

Stochastic Optimization Large-Scale Genomics

Development of scalable stochastic optimization algorithms for training deep models on terabyte-scale genomic datasets.

Interpretable Decision Trees Variant Pathogenicity

Development of interpretable machine learning models to classify genetic variant pathogenicity with human-readable decision rules.

Mixture Density Networks Uncertainty Quantification

Application of mixture density networks to quantify aleatoric and epistemic uncertainty in genomic predictions and biomarker discovery.

Biological Network Motif Detection Algorithms

Development of deep learning algorithms to identify recurrent network motifs and functional modules in biological interaction networks.

Protein Function Transfer Learning Homology

Application of transfer learning to predict protein function by leveraging evolutionary relationships and sequence homology information.

Clinical Trial Outcome Prediction Machine Learning

Development of machine learning models to predict patient response and clinical trial outcomes from baseline genomic and phenotypic features.

Recurrent Neural Networks Metabolic Pathway Modeling

Develops RNN architectures to capture temporal dynamics and sequential dependencies in metabolic pathway flux predictions and metabolite interactions.

Variational Autoencoders Single Cell Transcriptomics

Applies VAE frameworks to learn latent representations of single-cell gene expression data for cell type discovery and trajectory inference.

Language Models Protein Function Annotation

Leverages large pre-trained language models to predict and assign biological functions to uncharacterized proteins through sequence understanding.

Disentangled Representations Drug Property Prediction

Learns factorized latent factors to separately represent chemical, pharmacological, and toxicological properties of drug molecules.

Message Passing Neural Networks Chemical Reactions

Applies message passing frameworks to model reaction mechanisms and predict reaction products in synthetic biology workflows.

Normalizing Flows Variant Effect Distributions

Uses invertible neural networks to model complex distributions of genetic variant functional effects across populations.

Subgraph Sampling Methods Large Interaction Networks

Develops efficient sampling strategies to scale graph neural network inference on massive protein-protein interaction networks.

Conditional Generation Antibody Design Optimization

Creates conditional generative models to design antibodies with specified binding affinities and neutralization properties.

Physics-informed Neural Networks Molecular Dynamics

Incorporates physical constraints and conservation laws into neural networks for accurate molecular dynamics simulation acceleration.

Multi-task Learning Phenotype Genotype Correlation

Jointly learns multiple phenotypic prediction tasks to improve genotype-phenotype association discovery and generalization.

Ordinal Regression Disease Severity Stratification

Applies ordinal regression techniques to predict disease severity stages from genomic and clinical data while preserving ordinal structure.

Contrastive Divergence Sequence Motif Discovery

Uses contrastive learning to identify conserved DNA and RNA motifs through comparison of binding and non-binding sequences.

Hierarchical Clustering Latent Representations Omics

Combines hierarchical clustering with learned latent representations to discover multi-scale biological patterns in omics data.

Optimal Transport Cellular Trajectory Alignment

Applies optimal transport theory to align and compare single-cell developmental trajectories across different conditions and organisms.

Sparse Attention Mechanisms Long Sequence Genomics

Develops sparse attention patterns to efficiently process whole-genome sequences and detect long-range regulatory interactions.

Point Cloud Analysis Protein Surface Features

Represents protein surfaces as point clouds and applies 3D deep learning to extract binding site and function-relevant features.

Imbalanced Classification Rare Mutation Prediction

Addresses extreme class imbalance in predicting pathogenic effects of ultra-rare genetic variants through specialized loss functions.

Attention Visualization Gene Regulatory Logic

Uses attention weight visualization to interpret neural network decisions and discover interpretable gene regulatory logic rules.

Manifold Learning Single Nucleus Chromatin Accessibility

Applies nonlinear manifold learning to uncover intrinsic dimensionality and structure in single-nucleus ATAC-seq data.

Reinforcement Learning Synthetic Gene Circuit Design

Formulates gene circuit design as a reinforcement learning problem where agents optimize circuits for desired biological behaviors.

Sequence-to-sequence Models Protein Translation Prediction

Applies encoder-decoder architectures to predict codon usage patterns and translation efficiency from nucleotide sequences.

Mixture Model Clustering Cancer Subtype Discovery

Uses probabilistic mixture models to identify distinct cancer subtypes with prognostic relevance from multi-omics data.

Influence Functions Neural Network Prediction Robustness

Applies influence function theory to identify training samples most influential for biomedical predictions and improve robustness.

Generative Adversarial Networks Synthetic Patient Data

Develops GANs to generate realistic synthetic electronic health records and genomic data while preserving privacy.

Kernel Methods Epigenetic Mark Integration Analysis

Combines multiple kernel methods to integrate diverse epigenetic mark datasets for chromatin state prediction.

Longitudinal Analysis Microbiome Temporal Dynamics

Develops temporal models to track microbiome compositional changes and predict disease-associated dysbiosis patterns.

Edge Computing Mobile Genomic Analysis Applications

Creates efficient lightweight neural networks deployable on mobile and edge devices for point-of-care genomic analysis.

Heterogeneous Graph Learning Multi-source Biomedical Integration

Designs heterogeneous graph neural networks to integrate proteins, drugs, genes, and diseases from multiple biomedical databases.

Variational Inference Gene Expression Deconvolution

Applies variational Bayes methods to decompose bulk tissue gene expression into cell type-specific signatures.

Mutual Information Maximization Feature Selection Genomics

Uses information-theoretic approaches to select most predictive genomic features while minimizing redundancy.

Siamese Networks Biomarker Similarity Learning

Trains Siamese architectures to learn meaningful distance metrics between biomarkers for patient stratification.

Biological Network Pruning Model Interpretation

Develops pruning algorithms that maintain predictive performance while removing non-essential connections for biological insight.

Adversarial Examples Biomedical Model Vulnerability Assessment

Studies adversarial perturbations in genomic and imaging data to assess robustness of clinical AI systems.

Attention Flow Analysis Epistatic Interaction Networks

Uses attention mechanisms to identify and visualize epistatic gene interactions from high-dimensional genetic data.

Anomaly Scoring Genomic Quality Control Automation

Implements unsupervised anomaly detection to automatically flag low-quality genomic samples and sequencing artifacts.

Cross-modal Retrieval Molecular Imaging Data Fusion

Develops cross-modal learning to retrieve and align molecular images with corresponding structural and sequence data.

Continuous Bag of Words Biological Sequence Embedding

Applies CBOW-inspired methods to learn unsupervised embeddings of DNA and protein sequences for similarity search.

Stochastic Variational Inference Large Cohort Modeling

Implements scalable variational inference for Bayesian models on biobanks with millions of individuals.

Kernel Density Estimation Mutation Hotspot Identification

Uses kernel density estimation to identify statistically significant mutation clusters in genomic sequences.

Recurrent Dropout Temporal Patient Outcome Prediction

Applies specialized dropout regularization to RNNs for robust prediction of patient outcomes from clinical timeseries.

Attention Pooling Molecular Fingerprint Aggregation

Uses learned attention weights to aggregate molecular fingerprints for improved compound similarity assessment.

Multi-instance Learning Pathogenic Variant Detection

Formulates pathogenic variant identification as a multi-instance learning problem where variants within genes are instances.

Collaborative Filtering Drug Target Interaction Prediction

Applies collaborative filtering techniques to predict unknown drug-target interactions from sparse experimental data.

Attention Mechanism Interpretability SNP Importance Ranking

Leverages attention weights to automatically rank SNP importance for complex trait prediction models.

Graph Sparsification Biological Network Scalability

Develops graph sparsification techniques to maintain information while reducing computational complexity for biological networks.

Causal Mediation Analysis Gene Regulatory Pathways

Applies causal inference methods to decompose total genetic effects into direct and pathway-mediated components.

Multitask Benchmark Learning Protein Representation

Trains unified protein embeddings on multiple functional prediction benchmarks for improved generalization.

Explainable Clustering Disease Subtype Characterization

Combines clustering with interpretability techniques to characterize disease subtypes with clinically actionable biomarkers.

Neural Architecture Search Genomic Prediction Tasks

Automates neural network design discovery optimized specifically for genomic prediction and classification tasks.

Weighted Graph Isomorphism Conserved Protein Motif Alignment

Uses graph isomorphism techniques to align and discover conserved structural motifs across protein families.

Recurrent Neural Networks for Temporal Mutation Tracking

Develops RNN architectures to model time-dependent genomic mutations and predict evolutionary trajectories in cancer and viral populations.

Variational Autoencoders for Single-Cell Gene Expression

Applies VAE frameworks to disentangle and reconstruct high-dimensional single-cell transcriptomic data with interpretable latent representations.

Graph Isomorphism Networks for Protein Function Prediction

Leverages GIN architectures to classify protein functions by capturing structural patterns invariant to isomorphic transformations.

Persistent Homology for Biomolecular Structure Analysis

Applies topological data analysis techniques to extract robust topological features from protein conformations and molecular complexes.

Probabilistic Programming for Pathway Uncertainty Quantification

Develops probabilistic models to quantify uncertainty in biological pathway inference and gene regulatory network predictions.

Attention-based Sequence-to-Sequence CRISPR Guide Design

Creates seq2seq models with attention mechanisms to design optimal CRISPR guide RNAs minimizing off-target effects.

Normalizing Flows for Molecular Property Distribution Learning

Uses normalizing flow models to learn and sample from complex molecular property distributions for inverse design tasks.

Hierarchical Clustering with Neural Embeddings for Cell Types

Integrates hierarchical clustering with learned neural embeddings to discover cell type hierarchies in single-cell data.

Positional Encoding for Mutation Context Representation

Designs novel positional encoding schemes to represent genomic context around mutations for improved pathogenicity prediction.

Mutual Information Maximization for Omics Feature Selection

Applies information-theoretic approaches to identify minimal sufficient feature sets across proteomics and metabolomics data.

Epistasis Learning through Tensor Decomposition

Uses tensor decomposition methods to discover and model high-order genetic epistatic interactions from sequence variants.

Message Passing Neural Networks for Enzyme Classification

Develops message passing frameworks to classify enzyme function and substrate specificity from molecular graphs.

Optimization-based Neural Networks for Drug Binding Kinetics

Builds neural ODE models constrained by biochemical principles to predict drug association and dissociation rates.

Sparse Tensor Factorization for Multi-omics Integration

Applies sparse tensor factorization to identify latent factors explaining variation across genomics, proteomics, and metabolomics.

Contrastive Learning for Homologous Protein Discovery

Uses contrastive loss functions to learn protein representations that cluster homologs while separating distant relatives.

Implicit Neural Representations for Conformational Ensembles

Models protein conformational ensembles as implicit functions to efficiently represent and sample from continuous state spaces.

Adversarial Domain Adaptation for Batch Effect Correction

Employs adversarial learning to harmonize gene expression data across experimental batches and sequencing platforms.

Neural Rendering for Cryo-EM Density Map Interpretation

Applies neural rendering techniques to convert cryo-EM density maps into interpretable 3D protein structure models.

Mixture Density Networks for Phenotype Distribution Estimation

Uses MDN architectures to estimate multimodal distributions of phenotypic outcomes from genotypic and environmental data.

Causal Graph Learning from Perturbation Experiments

Develops methods to infer causal gene regulatory networks from CRISPR and chemical perturbation data.

Federated Transfer Learning for Decentralized Clinical Genomics

Combines federated learning with transfer learning to train predictive models across distributed healthcare institutions.

Set-based Neural Networks for Unordered Molecular Features

Develops permutation-invariant architectures to process unordered sets of molecular descriptors and atomic features.

Gradient-based Interpretability for Sequence-Function Models

Applies gradient saliency methods to identify critical sequence positions influencing predicted protein or RNA functions.

Neural Architecture Search for Biomedical Image Analysis

Uses AutoML techniques to automatically discover optimal neural architectures for histopathology and medical imaging tasks.

Symbolic Regression for Gene-Phenotype Relationship Discovery

Combines symbolic regression with neural networks to identify interpretable mathematical relationships between genes and phenotypes.

Heterogeneous Graph Neural Networks for Biomedical Knowledge Integration

Develops heterogeneous GNNs to integrate multiple node and edge types from biomedical knowledge graphs and databases.

Diffusion-based Sequence Alignment and Scoring

Applies diffusion models as a generative framework for sequence alignment and similarity scoring tasks.

Influence Functions for Identifying Critical Training Data

Uses influence functions to trace model predictions back to training examples and identify data points most influential for genomic models.

Equivariant Graph Neural Networks for Molecular Symmetry

Designs equivariant architectures that respect molecular symmetries and geometric transformations in drug discovery.

Recurrent Attention Mechanisms for Medical Report Generation

Develops attention-based models to automatically generate clinical interpretation reports from genomic sequencing data.

Spectral Methods for Protein Folding Simulation

Applies spectral neural network methods to model protein folding dynamics as eigenfunction decompositions.

Parametric and Non-parametric Mixture Models for Cell States

Combines parametric and non-parametric Bayesian approaches to discover discrete and continuous cell states from omics data.

Prototypical Networks for Few-shot Biomarker Discovery

Applies prototypical network framework to discover disease biomarkers from limited labeled patient samples.

Neural Exponential Families for Conditional Density Modeling

Uses neural exponential family models to capture complex conditional distributions in quantitative trait prediction.

Contrastive Divergence for Restricted Boltzmann Protein Models

Applies contrastive divergence learning to train RBM models capturing protein sequence coevolution patterns.

Adaptive Sampling Strategies for Drug Candidate Screening

Develops adaptive sampling algorithms combining uncertainty estimation and utility functions for efficient drug screening workflows.

Latent Factor Models for Phenotype-Genotype Decomposition

Learns latent factors that jointly explain variation in phenotypes and genotypes across populations.

Optimal Transport for Single-Cell Trajectory Inference

Leverages optimal transport theory to reconstruct developmental trajectories and transition probabilities between cell states.

Neural Tangent Kernels for Genomic Regression Analysis

Applies neural tangent kernel theory to provide theoretical insights into neural network behavior for genomic prediction.

Variational Inference for Bayesian Gene Network Reconstruction

Uses variational inference to efficiently approximate posterior distributions in Bayesian gene regulatory network models.

Attention Pooling for Hierarchical Representation Learning

Designs attention-based pooling mechanisms to aggregate information across molecular hierarchy levels for property prediction.

Physics-informed Neural Networks for Metabolite Kinetics

Incorporates biochemical rate equation constraints directly into neural networks for metabolite concentration prediction.

Uncertainty-aware Active Learning for Variant Curation

Combines uncertainty estimation with active learning to prioritize genomic variants for expert clinical review and curation.

Multiplex Network Analysis of Disease-Gene Associations

Models disease-gene relationships as multiplex networks integrating protein interaction, genetic, and phenotypic layers.

Submodular Optimization for Feature Importance in Genomics

Applies submodular optimization to identify minimal gene sets that preserve predictive performance for disease classification.

Siamese Networks for Protein Similarity and Clustering

Trains siamese neural networks to learn protein similarity metrics for unsupervised clustering and homology detection.

Stochastic Differential Equations for Gene Expression Dynamics

Models intrinsic and extrinsic noise in gene expression using neural SDE frameworks for dynamic prediction.

Gromov-Wasserstein Distance for Omics Data Alignment

Applies Gromov-Wasserstein optimal transport distances to align and compare omics datasets across different measurement modalities.

Geometric Deep Learning for Molecular Conformational Ensembles

This research develops geometric neural network architectures that leverage invariant and equivariant principles to model and predict the conformational landscapes and dynamic properties of biomolecules under physiological conditions.

Neuro-symbolic Integration for Systems Biology Modeling

This research combines neural network learning with symbolic reasoning and logic programming to construct interpretable, mechanistic models of complex cellular signaling pathways and metabolic networks from multi-omics data.