Imoto, H. | bioRxiv Science

bioRxiv2026

A genome-wide catalog of sites where peripheral DNA methylation reflects the brain

Epigenetic studies of psychiatric and neurological disorders rely largely on peripheral tissues, yet where peripheral DNA methylation (DNAm) reflects the brain remains unclear. We generated a paired brain and peripheral DNAm dataset from 148 donors in three cohorts and combined it with five published cohorts for consensus brain-peripheral correlations. By cross-validation across cohorts, we defined brain epigenome shared sites (BrESS), more than 100,000 CpGs per tissue at which peripheral DNAm reproducibly predicts brain DNAm. BrESS lay at promoters active in both brain and periphery and in housekeeping genes, and most were not explained by genetic effects. BrESS did not contribute to the heritability of seven disorders, and CpGs from epigenome-wide association studies of these disorders were not enriched at BrESS. Peripheral DNAm reflects the brain mainly in a regulatory compartment shared across tissues, outside which disease signals largely lie, and BrESS separate peripheral signals attributable to the brain from others.

neuroscience↗

bioRxiv2025

A structure-based modelling approach identifies effective drug combinations for RAS-mutant acute myeloid leukemia

Mutations activating RAS/RAF/MEK/ERK signaling are associated with poor outcome in acute myeloid leukemia (AML), but therapeutic targeting of this pathway is challenging. Here, we employ a structure-based, dynamic RAS pathway model to successfully predict RAF inhibitor (RAFi) combinations which synergistically suppress ERK signaling in RAS-mutant AML. Our in silico models predicted therapeutic synergy of two iterations of conformation-specific RAF inhibitors: Type I[1/2] + Type II and Type I + Type II. Predictions were validated in vitro in AML cell lines and patient samples, with synergy verified by the Loewe Additivity model. Lifirafenib (Type II) + encorafenib (Type I[1/2]) was highly synergistic against both NRAS- and KRAS-mutant lines, while synergy of lifirafenib + SB590885 (Type I) was specific to NRAS-mutants. Immunoblotting confirmed that combination efficacy correlated strongly with decreased RAS pathway activation. Leveraging the pharmacokinetic predictions of our in silico model, both combinations were then assessed in a pre-clinical NRAS-mutant AML patient-derived xenograft (PDX) model, showing significantly improved leukaemia growth delay and event-free survival compared with single agent approaches. Assessment of leukemia burden in bone marrow and spleen during treatment further showed site-specific efficacy against circulating and spleen-resident blasts for both combinations. In summary, we report that our structure based-modelling approach can effectively identify novel, non-obvious, and well-tolerated RAFi combinations that are highly effective against in vitro and in vivo models, thereby suggesting alternative potential therapeutic strategies for high-risk RAS-mutant AML.

cancer biology↗