Network Bio Launches with $50M to Build First Cell-Free RNA Foundation Model with NVIDIA
核心洞察
Network Bio (搜索) launched with $50 million in financing and announced a collaboration with NVIDIA (搜索) to build the world's first foundation model trained on cell-free RNA (搜索).
The company has assembled a research network of leading US academic biobanks providing access to 500,000+ patients across oncology, immunology, and metabolic and cardiovascular disease.
Network Bio (搜索)'s prior models, Exai-1 and Orion, detected early-stage lung cancer (搜索) at 94% sensitivity and 87% specificity, published in Nature Machine Intelligence and Nature Communications.
Palo Alto, CA-based biotechnology company Network Bio (搜索) launched today with $50 million in financing and announced a collaboration with NVIDIA (搜索) to build the world's first foundation model trained on cell-free RNA (搜索) (cfRNA). The collaboration will combine Network Bio's datasets from patient-derived tissues and paired blood samples linked to longitudinal clinical outcomes with NVIDIA's accelerated computing, NVIDIA Parabricks, and NVIDIA BioNeMo Recipes to advance disease detection, biomarker discovery, and drug development.
Network Bio (搜索) has built a research network of leading US academic biobanks, including Mass General Brigham, University of Pennsylvania, Duke University, and University of Colorado Anschutz, providing access to 500,000+ patients across oncology, immunology, and metabolic and cardiovascular disease. The network combines patient-derived tissue, paired blood samples, and longitudinal clinical outcomes, creating what Network Bio describes as the world's largest patient tissue training dataset of its kind.
A Foundation Model for Cell-Free RNA
Cell-free RNA (搜索) is among the richest sources of biological information in human blood, reflecting active gene expression across tissues in real time rather than static genomic sequence. That richness also presents a computational challenge: a single blood draw yields hundreds of millions of transcript-level observations, and the signal distinguishing early disease from normal variation is distributed across the full transcriptome rather than concentrated in a handful of markers.
Network Bio (搜索)'s prior work established that transformer-based models can learn this structure directly. The company's models, Exai-1, published in Nature Machine Intelligence, and Orion, published in Nature Communications, detected early-stage lung cancer (搜索) at 94% sensitivity and 87% specificity. Through the NVIDIA (搜索) collaboration, Network Bio will leverage NVIDIA's AI infrastructure to scale that approach from research cohorts to population-scale training, producing a foundation model capable of learning the semantic language of circulating RNAs that will be integrated with the company's Bio-Native AI platform.
"AI is poised to fundamentally transform medicine, but its success depends on access to high-quality biological data," said Asad Ali Ahmad, Ph.D., CEO and Co-Founder of Network Bio (搜索). "By combining NVIDIA (搜索)'s world-class AI expertise with Network Bio's tissue datasets and Bio-Native AI platform, we're creating the first foundation model for cell-free RNA (搜索). We believe this collaboration will unlock entirely new opportunities for earlier disease detection, biomarker discovery and the development of more precise therapies."
Collaboration Projects
Network Bio (搜索) and NVIDIA (搜索) announced two projects as part of the collaboration. The first is a population-scale cfRNA foundation model, in which the companies are collaborating to scale training of Nexus, a self-supervised transformer trained on cfRNA expression profiles, using NVIDIA BioNeMo Recipes including Transformer Engine. Early engineering milestones show an improvement in training throughput and reduction in time-to-convergence. The resulting model serves as the substrate for downstream supervised models spanning oncology and non-oncology indications.
The second project focuses on accelerated cfRNA bioinformatics with NVIDIA (搜索) Parabricks. cfRNA secondary analysis, alignment, quantification, and quality control across every sample entering the training set represents the throughput ceiling on how fast Network Bio (搜索) can convert biobank access into model-ready data. Integration of NVIDIA Parabricks has reduced per-sample processing time and compute cost, compressing the interval between sample acquisition and model training.
Vertically Integrated Platform
Network Bio (搜索)'s vertically integrated platform combines two foundational technologies: a biological research network and bio-native AI architecture purpose-built to learn from multimodal biological data, overcome technical confounders, and generate interpretable representations of disease biology that generalise across diseases. Peer-reviewed research on this approach has been published in top-tier journals, including Nature Machine Intelligence.
Following data harmonisation and integrated analyses, generated data is returned to commercial partners and its collaborating institutions, supporting basic science and biomedical research at participating US academic medical centres.
"From day one, we've been relentlessly focused on two parallel challenges — generating the large-scale biological datasets needed to explain human disease and building the AI systems capable of interpreting them," said Hani Goodarzi, Ph.D., and Raphael Potter, co-founders of Network Bio (搜索). "This collaboration with NVIDIA (搜索) represents an important step toward realizing General Medical Intelligence and transforming how new medicines are discovered."
Financing and Commercial Traction
The $50 million financing comes from Section 32, Thiel Bio, Founders Fund, Breyer Capital, Blue Venture Fund, JSL Health Capital, and other life sciences and AI investors. Network Bio (搜索) has already signed a $30 million-plus co-development and licensing agreement with a Fortune 100, top-10 healthcare company to develop next-generation AI models.
"Every patient leaves a barcode of their disease in their tissue and, until now, no one has been able to read those barcodes at scale," said Asad Ali Ahmad, CEO and co-founder of Network Bio (搜索). "This is the future of medicine."
The collaboration builds on Network Bio (搜索)'s mission to create General Medical Intelligence by training AI directly on large-scale tissue, blood, molecular, and clinical datasets. By combining multimodal biological data with state-of-the-art foundation models, the company is creating AI systems capable of understanding disease mechanisms across therapeutic areas and accelerating the development of novel diagnostics and therapeutics.