Shuai Cheng Li

This study shows that catalytic rewiring of Fanzor2's RuvC-II site activates trans-cleavage, enabling new diagnostic applications.

This paper introduces CAML, a deep learning model that improves protein function prediction by integrating sequence, structure, and contig-level CDS topology features, achieving significant performance gains over existing methods.

The paper introduces SpaMV, a method for integrating spatial multi-omics data that preserves modality-specific information and improves clustering and dimensionality reduction.

SpecImmune uses long-read sequencing to accurately genotype immune-related genes, showing improved accuracy over existing methods and revealing new genetic patterns.

TransHLA predicts epitopes across all HLA alleles with high accuracy, outperforming existing models in specificity and sensitivity for immunogenic and neoepitope identification.

The paper analyzes genome-wide coverage differences between nuclear DNA and cell-free DNA in healthy and cancer samples, identifying factors influencing coverage and developing a new cancer screening method.

SCOIT, a probabilistic tensor decomposition framework, improves latent embeddings from single-cell multiomic data, enhancing clustering and cross-omics analysis.

This paper introduces TCRanno, a tool to quantify TCR repertoire specificity, revealing insights into infections, autoimmunity, and cancer through analysis of 4,195 repertoires.

TCR2vec is a deep learning framework that learns representations of full T-cell receptors, outperforming methods that focus only on CDR3 regions in predicting binding specificity and clustering.

TEINet predicts TCR-epitope binding specificity using a deep learning framework with transfer learning, achieving an AUROC of 0.760 and outperforming baselines.

This paper introduces TCRpeg, a deep autoregressive generative model for TCR repertoires, achieving higher accuracy in probability estimation and improving multiple TCR-related tasks.

The paper introduces SEAT, a method that incorporates cell hierarchy to analyze functional diversity in single-cell data, improving clustering accuracy and revealing insights into cell signaling and regulatory networks.

GutMeta provides an online platform for analyzing and visualizing human gut metagenomic data with a curated database and analysis tools.

This paper introduces a graph convolutional network for analyzing nucleic acid sequences, outperforming existing methods in phage-related tasks.

This study expands the number of complete temperate phage genomes by over 100-fold using a novel computational method applied to bacterial NGS data.

The paper introduces PStrain-tracer, a software package to track microbial strain engraftment during FMT using shotgun metagenomic data, revealing strain-specific engraftment patterns and failure determinants.

SMURF embeds scRNA-seq data using matrix factorization with a Poisson-Gamma divergence objective, preserving self-consistency and recovering gene expression and cell cycle time courses.

SeCNV improves copy number profiling in single-cell DNA sequencing by using structural entropy, achieving high accuracy and efficiency on large datasets.

STORM reconstructs single-cell resolution quasi-structures from spatial transcriptomes by reducing pseudo affinities and embedding interactions into a low-dimensional space.

This study uses simulation and sequencing data to show that multiple SARS-CoV-2 variants can coexist in COVID-19 patients, with an increase in the number of coinfected variants over time.

SCYN is a single-cell CNV segmentation method using dynamic programming, showing high accuracy and speed on scDNA datasets.

The paper investigates allele-specific relationships between gene expression and genome structure in human cells using integrated haplotype-resolved data.

This paper introduces a conjugate graph model to reconstruct local genomic maps at oncovirus integration loci, revealing five affection patterns and mechanisms of viral integration in HPV and HBV-infected cancer samples.

scSVAS provides an online platform for real-time interactive visualization of single-cell copy number variation data, enabling efficient analysis and sharing of large-scale genomics results.

CNAHap phases germline variants using tumor allele-specific copy number alterations, achieving accurate haplotyping and long phasing blocks in Hepatocellular carcinoma.

SpecHap is a diploid phasing algorithm using spectral graph theory, showing efficiency and accuracy across multiple sequencing technologies and large genomes.

I-Impute is a self-consistent imputation method for scRNA-seq data that outperforms existing tools in correlation and clustering accuracy on both simulated and real datasets.

CovProfile identifies viral genomic variations and gene expressions from SARS-CoV-2 Nanopore sequences, revealing multiple strains and high expression of viral genes in patient samples.

DeepMF improves subtype discovery and data recovery in cancer omics data by handling noise and missing values with a deep learning-based matrix factorization model.

D-GPM uses deep learning to predict promoter methylation levels from landmark gene methylation data, outperforming traditional methods in accuracy and correlation.