Yanlin Zhang

This paper introduces USM-AA, an alignment-aware genomic foundation model that improves long-range DNA modeling by incorporating reference genome coordinates and reverse-complement masked learning.

BioMaster automates complex bioinformatics workflows using a multi-agent system, improving accuracy, efficiency, and scalability across tasks like RNA-seq and Hi-C processing.

ContextTAD improves TAD calling from Hi-C contact maps by learning context-aware boundary evidence, achieving strong performance across multiple evaluation settings.

PopGenAgent streamlines population genetics analyses by organizing tasks into a structured workflow, enabling reproducible and transparent multi-step analyses with preserved intermediate results.

The paper formalizes scientific methodology as a protocol for AI research, showing it produces auditable, evidence-backed outputs across domains and recovers known biological signals in genomics.

GreS improves spatial domain identification in spatial transcriptomics by integrating gene semantic priors with spatial and expression-based cues, achieving better accuracy and biological interpretability.

Mut-BPE improves variant effect prediction by enhancing BPE tokenization with single-nucleotide resolution, showing better performance in AUROC and AUPRC across multiple datasets.

NyxBind improves TFBS prediction by using contrastive learning to enhance DNN representations, outperforming existing models across benchmarks.

The paper analyzes whole-genome data from wild boars to trace their demographic history and identify genetic signatures of local adaptation in Central Asia.

MutBERT improves genomic foundation models by using a probabilistic representation of SNPs, outperforming existing models on downstream tasks.

Polaris is a tool for chromatin loop annotation in bulk and single-cell Hi-C data that outperforms existing methods across different datasets and conditions.