Lu Zhang

This study compares statistical and machine learning methods for predicting complex diseases using genomic variants, finding statistical methods generally outperform machine learning and ensemble models like Super Learner perform best.

dynDeepDRIM uses deep learning to predict direct regulatory interactions from single-cell time-course gene expression data, outperforming existing methods.

This paper presents a high-quality chromosome-level genome assembly of the Japanese eel and identifies gene duplication events and chromosomal reorganization linked to evolutionary adaptation.

AquilaDeepFilter improves the precision of large deletion SV detection from linked-reads by using a deep learning ensemble, while maintaining high recall and outperforming existing methods.

MARS detects haplotype-resolved structural variants from linked-read assemblies using Mendelian inheritance and HiFi reads, achieving 73%-87% validation rates for indels.

DeepDRIM is a deep neural network that reconstructs cell-type-specific gene regulatory networks from single-cell RNA-seq data, outperforming existing methods and distinguishing functional modules in COVID-19 patients.

METAMVGL improves metagenomic contig binning by integrating assembly and paired-end graphs to enhance accuracy for short contigs and correct binning errors.

Aquila uses linked-read data to assemble diploid genomes and detect genetic variants, achieving high coverage and accurate phasing of SNPs, indels, and structural variants.

Bfimpute is a Bayesian factorization method for imputing single-cell RNA-seq data, achieving better accuracy than existing methods on simulated and real data.

Aquila_stLFR is a tool for diploid genome assembly and structural variant calling using stLFR linked-read data, achieving high sensitivity for deletions and specificity for insertions.

This study demonstrates that de novo assembly of 10x linked-read data can effectively detect structural variants in personal genomes with high accuracy and cost-effectiveness.

This study evaluates how 10x Linked-Read sequencing parameters affect human genome assembly quality using real and simulated data, finding optimal coverage and fragment length ranges.

This paper introduces a multimodal framework to detect gene-gene interactions from large expression data, showing improved performance over existing methods in simulations and biological applications.

HAPDeNovo filters and phases de novo mutations in linked read sequencing data to reduce false positives while maintaining sensitivity.