Fig. 2
From: Predicting tumour content of liquid biopsies from cell-free DNA
Assessment of model predictions in a held-out test data set. a Fragment length distributions of cfDNA in a serial dilution of patient plasma into control plasma. b Predicted tumour content in the dilution series. The solid line corresponds to simple linear regression, showing that the model’s predictions correlate well with the concentration of patient-derived cfDNA. The dashed line is constrained to cross the origin, highlighting that our trained model overestimates tumour content in the low range