Steven L. Salzberg

Kmask reduces misclassifications in metagenomics by masking low-complexity regions, improving accuracy in bacterial, archaeal, viral, and fungal classification.

The paper identifies thousands of recursive splicing events in human introns using a novel bioinformatics pipeline applied to RNA-seq data from human brain gene expression studies.

T2T-HG002v1.1, a complete diploid HG002 assembly lacking detectable errors over 99.4% of its length, enables genome-based benchmarking of assemblies, sequencing reads, and phased variant calls.

EviAnn improves eukaryotic genome annotation by using transcript alignments and protein homology, outperforming existing tools with less computational time.

The paper tests AlphaFold2 and ColabFold's reliability by using false protein sequences and finds a low false positive rate, while also discovering new pseudogenes in the human genome.

The paper identifies off-target probe binding in 10x Genomics Xenium gene panels and develops a tool to assess its impact on spatial transcriptomic accuracy.

The paper presents HIV Atlas, a reference-quality annotation of HIV-1 and SIV transcriptional diversity, and demonstrates its utility in improving transcriptome analysis.

The paper analyzes microbial content in 5,734 whole-genome sequencing samples from TCGA to provide a comprehensive reference dataset for future cancer microbiome studies.

The paper identifies 584 recent intron gain events in human genes by comparing human and vertebrate genomes and analyzing intronization mechanisms.

The paper identifies isoswitching, a phenomenon of isoform usage changes in human brain development and aging, and shows it can predict age across the lifespan using RNA-seq data.

TD2 is a new tool for ab initio annotation of protein-coding ORFs in transcripts, showing sensitivity and precision across diverse eukaryotic transcriptomes.

OpenSpliceAI improves SpliceAI by implementing it in PyTorch, enabling faster processing and retraining on non-human species with reduced memory usage.

The paper assesses the conservation of human splice site annotations across 470 genomes, finding that MANE-annotated splice sites are highly conserved and better supported than others.

LiftOn improves genome annotation by integrating DNA and protein alignments, enabling accurate annotation transfer between closely and distantly related species.

PSAURON assesses protein-coding gene annotations across species using a machine learning model trained on over 1000 genomes, providing scores to identify high-quality annotations and potential spurious proteins.

The paper analyzes 2,199 upstream ORFs to determine if they can be novel exons added to existing genes, finding 582 with strong evidence for new protein isoforms.

Splam improves spliced alignments by accurately predicting splice junctions in DNA, outperforming existing methods like SpliceAI across human and non-human species.

WheresWalker is a mutation mapping pipeline that rapidly identifies mutation-containing intervals in whole genome sequencing data, validated in zebrafish and maize.

The paper applies a novel metagenomics analysis to SCTLD lesions and healthy tissues to identify disease-associated microbiome members.

The paper produces a high-quality genome sequence for whitebark pine and identifies NLR genes associated with resistance to white pine blister rust.

SPIT detects differential transcript usage in complex diseases by quantifying population heterogeneity and identifying subgroups with distinct DTU events, revealing new events in schizophrenia brain samples.

CHESS 3 improves upon previous versions by using better algorithms and filtering to create a more accurate catalog of human genes and transcripts.

The paper presents the complete sequence of a human Y chromosome, correcting errors in the GRCh38 reference and adding over 30 million base pairs.

ORFanage identifies ORFs in known and novel transcripts, improving annotations in RefSeq and GENCODE databases and enabling analysis of large RNA-seq datasets.

The paper assembles a gapless, reference-quality genome from a Southern Han Chinese individual and identifies gene-level differences compared to a reference genome.

KrakenUniq improves metagenomic classification on low-memory systems by processing databases in chunks without sacrificing accuracy.

The paper uses 3D protein structures to evaluate human protein isoforms for biological functionality, refining genome annotations.

The paper evaluates sequencing and assembly methods to create high-quality diploid human genome assemblies, finding that long reads and haplotype sorting improve accuracy and completeness.

JASPER improves genome assembly accuracy and creates population-specific references by polishing with k-mer counts and integrating variant data from multiple individuals.

This study presents a high-quality genome and methylome analysis of Quercus lobata, identifying genomic features linked to oak evolutionary success, including gene duplications and methylation patterns.