Gregory J. Zynda

A 40B-parameter DNA model trained on 9.3 trillion tokens scores coding and noncoding variants zero-shot, exposes interpretable sequence features, and writes chromosome-scale sequences.

This study combines high-throughput screening with deep learning to identify novel antibacterial compounds, achieving a 90X higher hit rate than traditional methods.

The paper compares DNA replication timing in mitotic and endocycling cells in maize roots, finding significant shifts in centromere replication timing during the endocycle.

Repliscan classifies replication timing regions from Repli-seq data by normalizing, filtering, and assigning replication signatures, outperforming previous methods with variable thresholds.