Fig. 2

Assure workflow and Assure Blood-based Cancer Detection AI development and validation. DNA and RNA are extracted from the subject’s plasma along with gDNA from the subject’s buffy coat. These analytes are then sequenced and processed through bioinformatic workflows to generate biologically relevant feature sets (Pillars). The pillars are fed into a panomic machine learning algorithm (ABCDai) to generate a prediction. The diagram also describes assay and AI development with samples used in each study. ABCDai-GPS was trained using samples from 660 patients with stage I and II cancers and 506 normal donors. Performance for ABCDai-GPS was evaluated using 5-fold cross validation.