Jianlin Jack Cheng

This paper introduces PLABench, a benchmarking framework for protein-ligand binding affinity prediction, revealing that structural input and training scale alone do not ensure cross-target generalization.

TransBind2 improves TF-DNA binding prediction by integrating TF structure, chromatin context, and bidirectional cross-attention, achieving high accuracy across human ChIP-seq data and cross-species predictions.

This paper reports on the fourth CAFA challenge, showing improvements in Gene Ontology term prediction but highlighting challenges in new evaluation settings.

This paper introduces FunBind, a multimodal model for protein function prediction that improves accuracy and enables zero-shot prediction of novel functions.

GRNPred is a graph transformer model pretrained on gene expression data to infer gene regulatory networks, achieving high AUROC and AUPRC scores on benchmark datasets.

This paper improves AlphaFold3 by using customized MSAs and templates, achieving higher accuracy in structure prediction for monomers, multimers, and protein-ligand complexes.

TomoSwin3D improves automated macromolecule detection in 3D cryo-ET tomograms by using a Swin Transformer-based U-Net with enhanced input features and occupancy-preserving supervision.

TFBindFormer improves TF-DNA binding prediction by integrating DNA and TF protein features using cross-attention, outperforming DNA-only models.

TransBind improves TF binding site prediction by integrating DNA sequences with protein embeddings, achieving higher accuracy and generalizability for new TFs.

GRNFormer infers gene regulatory networks from transcriptomics data using a graph transformer framework, outperforming existing methods and recovering known and novel regulatory interactions.

CryoVirusDB provides a labeled cryo-EM dataset for training AI to identify virus particles in micrographs, aiding in 3D structural modeling.

ETSAM is a fine-tuned AI method for segmenting cell membranes in cryo-ET tomograms, achieving state-of-the-art performance on an independent test set.

The paper improves AlphaFold2 and AlphaFold3 predictions using MSA engineering, extensive model sampling, and ensemble QA methods, achieving high accuracy in CASP16.

The paper introduces MICA, a deep learning method that combines cryo-EM density maps and AlphaFold3 predictions to improve protein structure modeling accuracy and completeness.

CryoFSL improves protein particle picking in cryo-EM micrographs with minimal annotations, achieving high accuracy and efficiency in low-SNR conditions.

The paper improves protein complex structure prediction by integrating AlphaFold2 and 3 with enhanced methods in MULTICOM4, achieving high TM-scores and DockQ scores in CASP16.

MPBind predicts protein binding sites for multiple partners using protein language models and equivariant graph neural networks, outperforming existing methods.

The paper presents a pipeline combining multi-modal deep learning and AlphaFold3 to improve atomic protein structure modeling from cryo-EM data.

CryoTEN improves cryo-EM density maps using a transformer-based model, achieving better quality and faster performance than existing methods.

This study overexpresses NPQ genes in C4 plant Setaria viridis to compare photoprotection mechanisms with C3 plants, finding shared and unique traits that could improve crop resilience.

The paper predicts protein complex stoichiometry by combining AlphaFold3 and template information, achieving high accuracy in CASP16.

The paper introduces a labeled dataset for AI-based cryo-EM map denoising, including experimental maps and three types of generated label maps for benchmarking.

GATE estimates protein complex model accuracy using graph transformers on pairwise similarity graphs, achieving high correlation and low ranking loss on CASP15 and CASP16 datasets.

This paper introduces TransPross, a transformer-based model for protein secondary structure prediction that outperforms existing methods by leveraging attention mechanisms and deep learning.

The paper uses multi-omics data and machine learning to analyze oviductal responses to gametes and embryos in mice, identifying region-specific transcriptional and proteomic changes.

Cryo2Struct improves de novo atomic protein structure modeling from cryo-EM density maps using a 3D transformer and HMM, outperforming Phenix on multiple metrics.

The paper introduces CryoSegNet, a method combining a specialized U-Net and SAM for cryo-EM particle picking, achieving higher accuracy than existing methods.

HiCDiff denoises single-cell Hi-C data using diffusion models, outperforming existing methods and working well on bulk Hi-C data.

The paper introduces Cryo2StructData, a large labeled cryo-EM density map dataset for training AI models to build atomic protein structures, showing it is larger and higher quality than existing datasets.

CryoTransformer improves protein particle picking in cryo-EM micrographs using a transformer model, achieving better resolution and F1-score than existing methods.