Balázs Ligeti

ProkBERT PhaStyle uses genomic language models to predict phage lifestyles, outperforming existing methods in accuracy and speed.

The paper introduces ProkBERT, a genomic language model family for microbiome analysis, showing strong performance in promoter prediction and phage identification.

CovidOutcome2 predicts disease severity from SARS-CoV-2 mutation profiles using autoML and deep neural networks, achieving ROC-AUC scores of 0.899 and 0.83 with and without patient age.

The paper uses machine learning to link SARS-CoV-2 mutations to COVID-19 severity, identifying 26 mutations and achieving high classification accuracy.

The paper analyzes the structural features of postsynaptic proteins to show that their combination of ordered and disordered regions enables diverse interactions crucial for learning and memory.