Natera announced a new multi-modal AI model designed to enhance its Signatera™ molecular residual disease (MRD) assessment by integrating longitudinal circulating tumor DNA (ctDNA) measurements, clinical records, digital pathology images and tumor sequencing data. Trained on outcomes from roughly 300,000 patients tested since Signatera’s launch, the model demonstrates statistically and clinically significant improvements in risk prediction, with sensitivity gains of 18% and specificity gains of 12% in preliminary validation. Data will be presented at an upcoming oncology conference ahead of planned rollouts in both research and clinical settings. Corporate Chief Medical Officer Dr. Alexey Aleshin highlighted that linking repeated ctDNA measurements with pathology, treatment regimens and patient outcomes over multiple years uniquely positions Natera to refine individual patient trajectories and guide personalized treatment decisions.

Natera entered a strategic collaboration with NVIDIA to leverage high-performance computing platforms, accelerated libraries and AI frameworks for training its next-generation multimodal foundation models. Under the agreement, Natera will integrate its proprietary longitudinal in vivo datasets and machine learning algorithms with NVIDIA Parabricks, BioNeMo and the NeMo Agent Toolkit. Early engineering benchmarks show a 75.7% reduction in genomic analysis run-time and a 59.2% cut in compute costs using Parabricks, as well as a 45.5% increase in model training speed with the Transformer Engine. Natera’s NeoSelect™ neoantigen prioritization algorithm and NeoPredict™ immunotherapy response predictor, which outperformed 25 established tools and delivered more than a two-fold improvement in hazard ratios versus tumor mutational burden, are among the first applications slated for scale-up. The partnership aims to build a computational foundation that spans precision oncology, organ health and women’s health by enabling real-time analysis of complex biological datasets at scale.