Patrick D. Hsu

Patrick Hsu's 2025 record reaches across two foundational layers of modern biomedicine: learning from genomes and rewriting them. As a co-senior author on Evo 2, he helped lead an openly released biological foundation model trained on more than 9.3 trillion nucleotides from over 128,000 genomes, with a one-million-nucleotide context window. The release included code, model weights and training resources, turning a large scientific asset into infrastructure that other laboratories could interrogate rather than a closed demonstration.

Later in the year, Hsu's group reported programmable bridge recombinases engineered for human cells. The system achieved insertions, excisions and inversions, with DNA segments mobilized up to roughly 0.93 megabases; reported insertion efficiency reached 20% in the tested setting, with genome-wide specificity as high as 82%. These are cell-based research results, not a therapy. Their importance is the size and programmability of the genomic changes, which address a class of variation that small-edit tools cannot readily model. Hsu's role is unusually legible: he is a senior author, laboratory leader and institution builder who has also argued for opening expensive scientific platforms to broad scrutiny.