Johannes Söding

PhyloScore improves RNA alignment covariation scores by incorporating phylogenetic trees, enhancing accuracy in predicting RNA secondary structures.

MMAD-Risk predicts age-dependent disease risks from plasma proteomes using a multivariate mixed AFT model, outperforming Cox regression on UK Biobank data.

PMGen improves peptide-MHC structure prediction and enables structure-guided peptide design, achieving high accuracy and enhancing ML model performance.

The paper introduces mwMCC, a method to detect multi-way chromatin interactions at sub-nucleosome resolution, revealing synergistic nano-scale interactions within super-enhancers and their role in gene regulation.

LoL-align improves protein structure alignment by maximizing a local log-odds score, outperforming TMalign and DALI in sensitivity and speed.

MAGmax improves genome recovery in metagenomic samples by merging and reassembling bins, producing higher-quality MAGs faster and with less memory than dRep.

Spacedust identifies conserved gene clusters in microbial genomes using structure comparison and statistical methods, recovering most known antiviral defense systems.

SoftAlign is an end-to-end protein structure alignment method that improves accuracy over existing tools while being compatible with other alignment strategies.

The paper benchmarks deep learning metagenome binners on CAMI2 datasets, finding COMEBin and GenomeFace as top performers, with post-binning reassembly improving low coverage bins.

Foldseek-Multimer rapidly aligns protein complexes by clustering superposition vectors, outperforming the gold standard in speed while maintaining alignment quality.

The paper identifies gut microbiome functional amyloids and links their abundance to neurodegenerative diseases like Parkinson's and Alzheimer's.

UniOP predicts operons in prokaryotic genomic and metagenomic data using intergenic distances, outperforming existing methods across multiple genomes and metagenomes.

PenguiN improves strain-resolved assembly of viral genomes and bacterial 16S rRNA from metagenomics by using overlap-based methods and a Bayesian model, resulting in longer contigs and more complete genomes.

CarpeDeam improves de novo assembly of ancient metagenomic datasets by incorporating damage-aware models, recovering longer sequences and proteins better than existing tools.

Foldseek's 3Di tertiary-interaction alphabet enables structure search four to five orders of magnitude faster than Dali and TM-align at 86% and 88% of their respective sensitivities.

The paper develops a thermodynamic model to predict binding affinity of multi-domain RNA-binding proteins, showing how cooperativity enhances specificity and affinity beyond individual domains.

The paper introduces Tejaas, a method for detecting trans-eQTLs by reverse regression, identifying 18851 unique trans-eQTLs across 49 tissues and showing enrichment in regulatory regions.

The paper introduces BMF, a thermodynamic model for discovering RNA-binding motifs that account for multivalent binding, showing bivalent binding is common and improves specificity.

BaMMmotif2 improves binding motif prediction by modeling higher-order dependencies, achieving higher recall and speed compared to existing tools.

SpacePHARER identifies phage-host relationships by matching CRISPR spacers to phage genomes in genomic or metagenomic data, improving sensitivity through protein-level comparisons and multiple match evidence.

MMseqs2 taxonomy assigns taxonomic labels to metagenomic contigs using fragment extraction and weighted voting, achieving 2-18x speed improvements over existing tools.

The paper analyzes Drosophila core promoter function by testing synthetic promoters with and without hormonal activation to understand how mutations affect gene expression.

The paper screens random peptides to identify transcription activation domains and develops a deep learning model to predict their function and characteristics.

MetaEuk annotates protein-coding genes in eukaryotic metagenomic data with high sensitivity and speed, discovering many novel proteins from under-sampled eukaryotic groups.

The paper introduces HH-suite3, an accelerated version of HH-suite for faster remote homology detection and protein annotation, with significant speed improvements over previous versions and other tools.

MERLoT reconstructs complex lineage trees from scRNA-seq data and imputes temporal gene expression profiles using simulated and real datasets.

The paper introduces Plass, a protein-level assembler that significantly improves protein sequence recovery from metagenomic samples.

PROSSTT simulates scRNA-seq data for complex differentiation processes with customizable lineage trees and noise, providing tools to evaluate lineage tree reconstruction.

The paper introduces a desktop and web server app for fast, interactive sequence searches using MMseqs2, achieving near-BLAST sensitivity with reduced response times.

This paper introduces B-LORE, a Bayesian multiple logistic regression method for GWAS, showing improvements in detecting disease-associated SNPs in heritable traits and unbalanced case-control ratios.