Zane Koch

Elucidating the structures of unknown small molecules remains a major challenge in chemistry, pharmacology, and drug discovery. While nuclear magnetic resonance (NMR) spectroscopy is the gold standard for structure elucidation, interpreting spectra remains labor-intensive and requires expert knowledge to navigate the vast space of possible structures. Here, we present an AI system for rapid, automated de novo structure elucidation that combines a molecular generative model with a scientific reasoning agent. Its generative component, C hef NMR2, is an atomic flow-matching model that predicts small-molecule structures from molecular formulae and one-dimensional ¹ H and ¹³ C NMR spectra. ChefNMR2 - Agen t augments our generative model with large language models (LLMs) to reason over and refine candidate structures. Across diverse benchmarks, ChefNMR2 substantially outperforms existing tools, offering calibrated confidence estimates and two to three orders of magnitude faster inference than frontier LLMs. We use ChefNMR2 - Agen t to elucidate the structures of four novel natural products, matching human-chemist accuracy in a fraction of the time. Our work demonstrates how specialized generative models and general-purpose reasoning agents can unite to solve challenging scientific inverse problems and accelerate NMR structure elucidation, thus opening new possibilities for AI-driven small-molecule discovery across diverse disciplines.

The DREAM complex has emerged as a central repressor of DNA repair, raising questions as to whether such repression exerts long-term effects on human health. Here we establish that DREAM-associated activity significantly impacts lifetime somatic mutation burden, and that such effects are linked to altered lifespan and age-related disease pathology. First, joint profiling of DREAM-associated activity (quantified from the expression of genes transcriptionally repressed by DREAM) and somatic mutations across a single-cell atlas of 21 mouse tissues shows that cellular niches with lower DREAM-associated activity have decreased mutation rates. Second, DREAM-associated activity predicts the varied lifespans observed across 92 mammals, with low activity marking longer-lived species. Third, reduced DREAM-associated activity in individuals with Alzheimer’s disease predicts late disease onset and decreased risk for severe neuropathology. Finally, DREAM knockout in mice protects against mutation accumulation, reducing single-base substitutions by 4.2% and insertion/deletions by 19.6% in the brain. These findings position DREAM as a key regulator of aging.

Despite the well-known rise in somatic mutations with age, their true impact on aging—and the origins of mosaicism—has remained elusive. Here we present a powerful new framework that reconstructs the timing of somatic mutations across the entire lifespan using only adult tissues, integrating single-cell transcriptomics with high-resolution variant calling. Applying this framework to human and mouse tissues, we find that the vast majority of somatic mutations do not accumulate gradually with age but instead arise in a dramatic burst during embryonic development, concentrated in the first weeks of life. These early mutations seed long-lived clonal lineages that span tissues and persist for decades, defining the mosaic architecture of the body far more than previously appreciated. Our findings overturn the conventional view of mutation accrual in aging and highlight prenatal development as the dominant origin of lifelong somatic mosaicism.

Epigenetic remodeling is a hallmark of aging, yet which epigenetic layers are most affected during aging—and the extent to which they are interrelated—is not well understood. Here, we perform a comprehensive analysis of epigenetic aging encompassing 6 histone marks and DNA methylation measured across 12 tissues from > 1000 humans and mice. We identify a synchronized pattern of age‐related changes across these epigenetic layers, with all changes converging upon a common set of genes. Notably, an epigenetic clock based on these genes can accurately predict age using data from any layer (Spearman ρ : 0.70 in humans, 0.81 in mice). Applying this “pan‐epigenetic” clock, we observe that histone modification and DNA methylation profiles agree in the prediction of which individuals are aging more rapidly or slowly. These results demonstrate that epigenetic modifications are subject to coordinated remodeling over the lifespan, offering a unified view of epigenetic aging.

Aging is a complex biological process characterized by molecular changes across multiple biological scales. While these alterations have been extensively studied in humans and rodents, the molecular changes associated with aging in dogs remain underexplored despite their relevance as a model for human aging. In this study, we profiled gene expression ( n = 16,273 genes) and protein abundance ( n = 2041 proteins) in whole blood and blood plasma from 40 laboratory beagles across young (3–5 years old, n = 10), old (8–9 years old, n = 17), and geriatric (10–14 years old, n = 13) life stages. We identified 816 genes and 40 proteins that significantly changed in abundance during aging, converging on pathways involved in DNA repair, collagen processing, and inflammation. Notably, these canine aging signatures overlapped with human age-associated genes, including those tied to the hallmarks of aging, reinforcing the existence of conserved aging mechanisms across species.

The DREAM complex has emerged as a central repressor of DNA repair, raising questions as to whether such repression exerts long-term effects on human health. Here we establish that DREAM activity significantly impacts lifetime somatic mutation burden, and that such effects are linked to altered lifespan and age-related disease pathology. First, joint profiling of DREAM activity and somatic mutations across a single-cell atlas of 21 mouse tissues shows that cellular niches with lower DREAM activity have decreased mutation rates. Second, DREAM activity predicts the varied lifespans observed across 92 mammals, with low activity marking longer-lived species. Third, reduced DREAM activity in Alzheimer’s patients predicts late disease onset and decreased risk for severe neuropathology. Finally, we show DREAM knockout protects against mutation accumulation in vivo , reducing single-base substitutions by 4.2% and insertion/deletions by 19.6% in brains of mice. These findings position DREAM as a key regulator of aging.

DNA methylation marks have recently been used to build models known as “epigenetic clocks” which predict calendar age. As methylation of cytosine promotes C-to-T mutations, we hypothesized that the methylation changes observed with age should reflect the accrual of somatic mutations, and the two should yield analogous aging estimates. In analysis of multimodal data from 9,331 human individuals, we find that CpG mutations indeed coincide with changes in methylation, not only at the mutated site but also with pervasive remodeling of the methylome out to ±10 kilobases. This one-to-many mapping enables mutation-based predictions of age that agree with epigenetic clocks, including which individuals are aging faster or slower than expected. Moreover, genomic loci where mutations accumulate with age also tend to have methylation patterns that are especially predictive of age. These results suggest a close coupling between the accumulation of sporadic somatic mutations and the widespread changes in methylation observed over the course of life.