Philipp Lorenz

The paper introduces BaseData, a new biological sequence database that expands known protein diversity by over 10x using a global biodiscovery pipeline.

An enzyme-specific conditional language model generates active carbonic anhydrases and lactate dehydrogenases to user specification, some under 40% identity to natural proteins, with no retraining.

The paper improves AlphaFold2 performance using a global metagenomic data supply chain, achieving better protein structure predictions on CASP15 and CAMEO targets.

The paper investigates the relationship between H3K4me3 and transcription, finding no direct instructive or informative link between the two on a genome-wide scale in budding yeast.