Dr. Rob Bradley from Fred Hutchinson Cancer Centre, the McIlwain Family Endowed Chair and Director of the Translational Data Science Integrated Research Centre, has led the creation of Generate Expression Model‑1 (GEM‑1), a generative genomics foundation model trained on the world’s most comprehensive annotated RNA‑seq dataset. Using this model, Synthesize Bio can generate in‑silico gene‑expression data that matches wet‑lab experiments directly from experimental design descriptions, offering a new paradigm for predicting experimental outcomes and accelerating drug discovery.
Synthesize Bio Secures 10 Million Seed Funding From Madrona And Partners In Seattle 2025
Synthesize Bio, a Seattle‑based biotechnology start‑up founded in 2025 by Dr Rob Bradley, PhD, and Dr Jeff Leek, PhD, announced that it has secured a $10 million seed investment led by Seattle‑based venture capital firm Madrona. The round also attracted capital from AI2 Incubator, Sahsen Ventures, Inner Loop Capital and Point Field Partners, underscoring broad confidence in the company’s generative genomics platform. The funding will accelerate the development of the company’s first‑generation foundation model, Generate Expression Model‑1 (GEM‑1), refine its curated RNA‑seq training data, and expand its partnership network with biopharmaceutical companies and academic researchers.
Dr Rob Bradley, PhD, McIlwain Family Endowed Chair and Director of the Translational Data Science Integrated Research Center at Fred Hutchinson Cancer Centre, and Dr Jeff Leek, PhD, J Orin Edson Foundation Endowed Chair and Chief Data Officer at the same institution, co‑founded Synthesize Bio in 2025. Their combined expertise in RNA informatics and translational data science underpins the company’s generative genomics platform. Leek’s work on RNA informatics culminated in the assembly, normalisation and federation of disparate RNA‑seq datasets from around the globe, producing the most extensive annotated RNA‑seq repository compiled to date, which serves as the training foundation for GEM‑1.
GEM1 Predicts In Silico Gene Expression Outcomes Matching Wet Lab Experiments Demonstrating New Paradigm For Drug Target Discovery
A preprint posted to bioRxiv formally documents that GEM‑1 can reproduce the outcomes of laboratory experiments conducted after its training data cutoff. The model ingests detailed experimental design specifications—cell type, perturbation, time‑point—and outputs simulated RNA‑seq data that matches wet‑lab measurements with unprecedented accuracy. In addition, GEM‑1 can generate synthetic expression data for large clinical cohorts, producing distributions that closely mirror real patient samples, thereby demonstrating its capacity to extrapolate beyond its training window and simulate clinical‑trial outcomes.
Synthesize Bio is forging formal collaborations with a range of biopharmaceutical companies to embed GEM‑1 into the early stages of drug development. The partnership framework is designed to de‑risk the entire clinical pipeline, beginning with the identification of high‑confidence therapeutic targets and extending through to the optimisation of trial design. By offering GEM‑1 through R and Python API clients, Synthesize Bio enables biopharma teams to integrate the model directly into their existing data‑analysis workflows, accelerating decision‑making and reducing reliance on costly wet‑lab experiments.
Synthesize Bio has launched programmatic interfaces that expose GEM‑1 to the wider scientific community. The new R and Python client libraries, released today, allow researchers to submit experimental design specifications and retrieve synthetic gene‑expression matrices directly from the model’s cloud‑based inference engine. This initiative democratises access to generative genomics, enabling laboratories and biopharma teams worldwide to embed GEM‑1 into their existing bioinformatics pipelines without specialised hardware or deep machine‑learning expertise.
Original Press Release
Source: View Original Source
