Rolf Backofen

This study evaluates AlphaFold 3's accuracy and reliability in predicting RNA structures and protein-RNA interfaces, finding that confidence scores do not always correlate with correct predictions.

CRISCross improves CRISPR-Cas9 off-target prediction by integrating epigenetic features with sequence data across cell types, showing better performance than sequence-only models.

PGRanker improves Mendelian gene prioritization by integrating phenotype data with gene networks using graph learning, achieving higher recall and better ranking of causal genes.

This study identifies 338 previously unannotated small open reading frames in UPEC strain CFT073, with functional evidence for nine intergenic sORFs and potential roles in stress response.

This paper revises the genome annotation of Synechocystis using ribosome profiling and proteogenomics, identifying new genes and correcting existing annotations.

The paper shows IL-1β enhances IGFL1 translation through a G-rich 3UTR element in MCF7 cells.

The paper identifies structural variants in MENA individuals using long-read sequencing, revealing previously unreported variants and population-specific disease-related variants.

This paper introduces a foundation model for CRISPR array detection that works without metagenomic assembly, achieving high accuracy on both short and long sequences.

The paper introduces GraphMeX-plain, a graph neural network with explainable AI to prioritize disease-specific CpG sites as potential biomarkers in breast cancer and acute myeloid leukemia.

The paper analyzes CRISPR systems in prokaryotes, finding that multiple CRISPR arrays co-occur with Cas loci, vary in structure, and may function independently or collaboratively.

The paper analyzes CRISPR array repeat mutations across prokaryotic genomes, revealing subtype-specific patterns and a link between repeat mutations and spacer dynamics.

CRISPR-evOr predicts CRISPR array orientation by analyzing evolutionary patterns, improving accuracy for arrays other tools cannot confidently resolve.

RAPDOR uses Jensen-Shannon distance to analyze complex proteomics datasets, identifying RNA-binding proteins in cyanobacteria and showing its utility in stress-induced protein redistribution.

The paper introduces AVDB, a database of genetic variants from Arab populations, providing pathogenicity classifications and disease burden metrics to improve genomic screening and diagnostics.

The paper investigates the structural determinants of inverted Alu-mediated backsplicing in the HIPK3 gene using SHAPE-JuMP and finds that long-range Alu-mediated base pairs are required for circularization, but not full-length Alu elements.

The paper evaluates discrepancies in dimensionality reduction techniques for time-series scRNA-seq data using synthetic data and finds that different methods may not reliably represent dynamics when used alone.

The paper identifies a new subgroup of type V-K CAST systems in cyanobacteria through phylogenetic analysis and characterizes its unique features.

This work introduces a dataset and a transformer-based model for classifying ncRNA-protein interactions, showing strong performance compared to existing methods.

The paper identifies Pantr1 as a lncRNA regulated by FOXG1 that facilitates FOXG1's site-specific chromatin binding and impacts neuronal development.

SpacerPlacer reconstructs ancestral CRISPR array evolution, revealing frequent spacer deletions and a decrease in deletion frequency towards both ends of arrays.

This paper introduces a new RNA design paradigm called partial RNA design and presents an algorithm, libLEARNA, that efficiently solves constraint RNA design tasks with variable lengths and precise control over sequence and structure constraints.

RNAformer predicts RNA secondary structure directly from sequence using a simple axial-attention model, achieving state-of-the-art performance while addressing homology concerns.

This paper develops a Transformer-based tool recommender system for Galaxy, achieving faster convergence and better generalization than a previous RNN-based system.

DEWSeq improves eCLIP data analysis by increasing the detection of RNA-binding protein binding sites and their functional elements.

The paper presents a docker-based Jupyterlab infrastructure in Galaxy for AI development, enabling remote training and management of AI projects with GPU support and version control.

The paper expands the known proteome of Sinorhizobium meliloti by identifying 48 novel small proteins using Ribo-seq and proteogenomics.

The paper uses complementary Ribo-seq methods to map the translatome of Campylobacter jejuni, identifying novel small open reading frames and refining gene annotations.

The paper introduces a chromatin loop detection algorithm using Hi-C data with strict candidate selection and statistical testing.

The paper benchmarks ORF prediction tools for bacterial Ribo-seq data, finding DeepRibo performs well but none detect all novel sORFs with high sensitivity.

StoatyDive evaluates and classifies peak profiles from sequencing data to improve downstream analysis and quality control.