DIGITAL BIOLOGY
Features
Weave Bio and Takeda expand AI-native regulatory automation across 14 active drug programs
The partnership now covers questions from health authorities and draft responses, alongside Investigational New Drug (IND) drafting. A 2025 study co-authored by both companies tested nonclinical summaries from two historical INDs. Initial drafts took 2.6 and 3.7 hours, versus an estimated 100 hours of manual drafting each. The study assessed initial drafts rather than submission-ready documents, and expert review found deficiencies that required human editing.
Owkin licenses its K Pro AI Scientist and oncology data to Boehringer Ingelheim
The collaboration builds on a 2025 pilot using Owkin's MOSAIC dataset to study the tumor microenvironment around a gene target. Owkin will also generate new multimodal immunology data, and Boehringer's teams will use K Pro to examine the datasets and run analyses. Owkin says the system can help scientists generate, test and prioritize hypotheses, with the aim of accelerating discovery across oncology and immunology.
Ono taps Aitia's causal AI digital twins to find neurology drug targets
Aitia will use its causal AI-based REFS technology to build Gemini Digital Twins, virtual patient models, of neurological diseases and identify new therapeutic targets and biomarkers. Ono has an exclusive worldwide option to research, develop and commercialize drug candidates against the targets Aitia identifies. The companies say REFS infers causal relationships to find targets that conventional correlation-based AI cannot identify, an approach they see as suited to neurology's complex genetic, developmental, neural network, immune and environmental interactions. Financial terms were not disclosed.
ai& and Tenstorrent launch JapanFold, a sovereign drug discovery AI platform
JapanFold serves open-source structural biology models, including Boltz-2, OpenFold3, RFdiffusion 3 and BoltzGen, running on Tenstorrent Galaxy superclusters entirely inside Japan. ai& says the platform matches conventional accuracy without a cloud cost premium, because output scales nearly linearly as more processors are added. Researchers can reach it through a browser workbench, a single API endpoint, or agent skills built for tools such as Claude Code and Cursor.
Insilico study finds lower aging-clock estimates in patients treated with rentosertib
Researchers applied six proteomic aging clocks to samples from 42 patients in an existing 12-week phase 2a trial of the AI-designed drug for idiopathic pulmonary fibrosis. The clocks predicted lower biological age in treated groups, although effects varied by dose and timepoint. This exploratory analysis cannot fully separate effects on aging from changes related to pulmonary fibrosis, and it does not establish age reversal or longer life.