Rob S Egan

This study evaluates FP8 quantization in GenomeOcean, finding it preserves biological fidelity while improving efficiency, but throughput gains fall short of theoretical limits.

The paper introduces GenomeOcean, a large-scale generative model trained on metagenomic data, to explore the genomic manifold hypothesis and demonstrates its alignment with another model.

This study evaluates using persistent memory to reduce out-of-memory errors in metagenome assembly, showing it can handle terabyte-sized datasets with increased scalability.

The paper evaluates metagenomic software tools using realistic datasets, highlighting improvements in assembly and taxon profiling while identifying remaining challenges in virus/archaea detection and reproducibility.

Genome Constellation improves taxa assignment and discovers novel taxa in metagenomic datasets by clustering genomes based on k-mer similarities.

MiniScrub improves Nanopore read quality through de novo error correction, leading to better genome assembly with fewer misassemblies and indel errors.

Base-by-base tests of raw nanopore current, native versus amplified DNA, detect 4mC, 5mC and 6mA without training data, recovering seven methylases' motifs (AUC 0.62–0.97).