Network Bio Raises $50 Million and Partners with NVIDIA to Build a Model Using Blood RNA

Network Bio Raises $50 Million and Partners with NVIDIA to Build a Model Using Blood RNA

On August 19, biotechnology company Network Bio announced $50 million in financing and a collaboration with NVIDIA. The company plans to train its Nexus model on linked tissue and blood samples, together with records of the patients’ subsequent health outcomes.

RNA helps cells read genetic instructions. Fragments of RNA enter the bloodstream, where they provide a record of which genes were active in different tissues when the sample was collected. Nexus is intended to identify recurring patterns in these profiles that are associated with the later development of disease.

For training, the company is using a network of biobanks, which store biological samples for research, to collect blood and tissue from the same patient together with information about what happened to that patient later. Network Bio says it is standardizing selection criteria, quality control, and data from different sites. This allows the molecular pattern in a sample to be compared with the patient’s subsequent disease course.

A spring War on the Rocks essay on biological data for AI described the same problem at the policy level: medical data must be compatible and verifiable if models are to use different datasets together. Network Bio is trying to assemble such a dataset for specific samples linked to patient histories.

“Every patient leaves a barcode of disease in their tissues, and until now, no one has been able to read those barcodes at this scale,” says Network Bio CEO and co-founder Asad Ali Ahmad.

Before training begins, NVIDIA Parabricks software will prepare the data from each sample by aligning the sequenced RNA fragments to the genome, quantifying them, and checking data quality. In the collaboration announcement, Network Bio describes Nexus as a model that learns from RNA profiles without predefined diagnoses and could later serve as a foundation for specific applications, including applications in oncology.

Network Bio names Mass General Brigham, the University of Pennsylvania, and the University of Colorado Anschutz among its biobank partners. Pharmaphorum reports that the network gives Network Bio access to data from more than 500 thousand patients.

The financing, biobank network, and data processing form a single workflow. A sample is collected together with the patient’s subsequent history, the data are standardized, and the resulting dataset is then used to train the model.