Eva Marsalkova

This study uses ESM3 to generate knotted proteins, finding high success rates with guided decoding and robustness in real knotted proteins.

miRBind2 is a deep learning model that predicts miRNA binding and transcript repression from sequence, outperforming existing methods with fewer parameters.

AlphaFind v2 enables fast, structure-based search across protein structures and domains, using embeddings and US-align for refinement and supporting organism and CATH-based filtering.

The paper compares AlphaFold2, ESMFold, and OmegaFold on new PDB structures, finding AlphaFold2 most accurate but slower, while ESMFold and OmegaFold offer faster predictions with varying strengths in specific cases.

The paper introduces miRBench, a benchmark dataset for microRNA binding site prediction that reduces miRNA frequency class bias, showing improved performance with a simple CNN model.

This study uses machine learning to predict protein knotting from amino acid sequences, achieving 98.5% accuracy using AlphaFold predictions and UniProt data.

The paper identifies AGO2 target sites using chimeric reads from CLASH and eCLIP experiments, revealing microRNA and non-microRNA drivers and developing predictive models and tools.

This study investigates the role of the SAP domain in Ku-DNA interaction phases, showing it facilitates initial binding but not sliding, and finds that reduced loading efficiency does not impair NHEJ or telomere maintenance in Arabidopsis.

PENGUINN uses Convolutional Neural Networks to predict G-quadruplex forming sequences, outperforming existing methods.