Yuedong Yang

Optimizing broadly neutralizing antibodies via all-atom interaction modeling and pre-trained language models

This paper introduces InterAb and InterAb-Opt, computational frameworks that predict and optimize broadly neutralizing antibodies using all-atom modeling and pre-trained language models, showing significant improvements in binding affinity and neutralization activity.

Bridging LLM Reasoning and Chemical Knowledge via an Evolutionary Multi-Agent Framework for Molecular Synthesis

This paper introduces EvoSyn, an evolutionary multi-agent framework that combines LLM reasoning with domain validation to generate chemically feasible molecules.

Accurate Protein-Protein Interactions Modeling through Physics-informed Geometric Invariant Learning

ProTact improves protein-protein interaction prediction using a geometric graph neural network with physics-informed constraints, outperforming existing methods on benchmark datasets.

Structural determinants of IGHV1-69 public antibodies conferring resilience to SARS-CoV-2 antigenic escape

This study identifies structural features of affinity-matured R1-32-like antibodies that enable tolerance to SARS-CoV-2 mutations and reveals an ultrapotent antibody with broad variant activity.

Advancing Protein Design via Multi-Agent Reinforcement Learning with Pareto-Based Collaborative Optimization

MAProt improves protein design by combining structure-based and language model approaches using multi-agent reinforcement learning with Pareto-based optimization, achieving better performance across multiple tasks.

AI recognizes convergent somatic hypermutation signatures to allow the discovery of variant-resilient broadly neutralizing antibodies

The paper uses AI to identify convergent somatic hypermutation patterns in antibodies, enabling the discovery of variant-resilient broadly neutralizing antibodies from early-pandemic data.

A synergistic generative-ranking framework for tailored design of therapeutic single-domain antibodies

This paper introduces TFDesign-sdAb, a deep learning framework for designing therapeutic single-domain antibodies with tailored functionalities, achieving high success rates in generating Protein A-binding sdAbs.

CellFM: a large-scale foundation model pre-trained on transcriptomics of 100 million human cells

CellFM is a large-scale foundation model pre-trained on 100 million human cells, outperforming existing models in cell annotation, perturbation prediction, and gene function prediction.

Interpretable Drug Response Prediction through Molecule Structure-aware and Knowledge-Guided Visible Neural Network

DrugVNN predicts drug responses with high accuracy and interpretability by integrating gene features and drug representations using a novel attention mechanism.

Evaluation and optimization of sequence-based gene regulatory deep learning models

The paper evaluates and optimizes sequence-based gene regulatory deep learning models using a DREAM Challenge dataset, showing improvements through architectural and training strategy refinements.

Cross-modal Graph Contrastive Learning with Cellular Images

The paper proposes a cross-modal graph contrastive learning method that uses cellular images to improve molecular representation learning for downstream tasks.

Genome-scale annotation of protein binding sites via language model and geometric deep learning

GPSite predicts protein binding sites across multiple ligands using a multi-task network, outperforming existing methods and enabling genome-scale annotations.

Predicting Disease-gene Associations through Self-supervised Mutual Infomax Graph Convolution Network

This paper introduces a graph convolutional network to predict disease-gene associations by leveraging external collaborative graphs and reducing noise through mutual information maximization.

DiffDec: Structure-Aware Scaffold Decoration with an End-to-End Diffusion Model

DiffDec generates structure-aware R-groups for molecular scaffolds using a diffusion model, achieving higher recovery rates and better binding affinity than existing methods.

SE(3) Equivalent Graph Attention Network as an Energy-Based Model for Protein Side Chain Conformation

This paper introduces GraphEBM, an SE(3)-equivariant graph attention network for protein side chain conformation, outperforming existing methods in rotamer recovery and combinatorial optimization.

Accurately identifying nucleic-acid-binding sites through geometric graph learning on language model predicted structures

GLMSite identifies DNA and RNA-binding sites using geometric graph learning on ESMFold-predicted structures, outperforming sequence-based methods and matching structure-based approaches.

Characterizing RNA-binding Ligands on Structures, Chemical Information, Binding Affinity and Drug-likeness

The paper introduces RNALID, a database of 358 RNA-ligand interactions validated by low-throughput experiments, showing novel ligands and structural patterns with implications for drug discovery.

Identifying B-cell epitopes using AlphaFold2 predicted structures and pretrained language model

GraphBepi predicts B-cell epitopes using AlphaFold2 structures and a pretrained language model, outperforming existing methods in AUC and AUPR.

Fast and accurate protein function prediction from sequence through pretrained language model and homology-based label diffusion

SPROF-GO predicts protein functions from sequences using a pretrained language model and label diffusion, outperforming existing methods by over 10% in AUPR.

Self-supervised learning on millions of pre-mRNA sequences improves sequence-based RNA splicing prediction

This study introduces SpliceBERT, a self-supervised pre-trained language model for RNA splicing prediction, showing improved performance in variant effect prediction and cross-species splice site identification.

A Meta-learning based Graph-Hierarchical Clustering Method for Single Cell RNA-Seq Data

MeHi-SCC improves single-cell RNA-seq clustering by integrating meta-learning and multi-dataset information to identify cell subtypes in large-scale atlases.

Deciphering Spatial Domains by Integrating Histopathological Image and Transcriptomics via Contrastive Learning

This paper introduces ConGI, a method that integrates gene expression and histopathological images via contrastive learning to improve spatial domain identification in spatial transcriptomics.

Alignment-free metal ion-binding site prediction from protein sequence through pretrained language model and multi-task learning

LMetalSite predicts metal ion-binding sites from protein sequences using a pretrained language model and multi-task learning, achieving higher accuracy than structure-based methods.

An uncertainty-based interpretable deep learning framework for breast cancer outcomes prediction

This paper introduces UISNet, a deep learning framework for breast cancer outcomes prediction that improves accuracy and interpretability by integrating uncertainty-based gradients and biological knowledge.

Spatial Transcriptomics Prediction from Histology jointly through Transformer and Graph Neural Networks

Hist2ST predicts RNA-seq expression from histology images using a deep learning model combining Transformer and graph neural networks, outperforming existing methods in gene expression prediction and spatial region identification.

A Parameter-free Deep Embedded Clustering Method for Single-cell RNA-seq Data

ADClust is a parameter-free deep embedding clustering method for scRNA-seq data that automatically determines the number of clusters and outperforms existing methods.

Capturing large genomic contexts for accurately predicting enhancer-promoter interactions

TransEPI improves enhancer-promoter interaction prediction by using a Transformer model to capture long-range genomic dependencies across different cell types and chromosomes.

A Robust and Scalable Graph Neural Network for Accurate Single Cell Classification

This paper introduces GraphCS, a scalable GNN method for single-cell classification, showing improved performance on large datasets.

AlphaFold2-aware protein-DNA binding site prediction using graph transformer

The paper introduces GraphSite, a method using AlphaFold2-predicted structures and graph transformers to predict DNA-binding residues, outperforming existing sequence and structure-based methods.

DeepANIS: Predicting antibody paratope from concatenated CDR sequences by integrating bidirectional long-short-term memory and transformer neural networks

DeepANIS predicts antibody paratopes by integrating BiLSTM and transformer networks on concatenated CDR sequences, outperforming existing methods.