Fig. 1

Non-haematological cell features are input/output models to simulate multicancer screening. A) Baseline percentage related to the clinical presentation of healthy subjects enrolled as controls (CS) and the type of tumours in the cancer patient group (CP). B) Different proportions of CD45 pos/ neg cells in liquid and solid matrices. In C) Identification of six patterns (Pn) by heatmap depicting expression levels of each marker on individual cell. Principal Analysis Component (PCA) maps allowed for a multidimensional separation of cell populations in CS and CP groups E) Cytological images of blood-derived specimens in CS (SUB109) and CP (PZ014) and immunofluorescence analysis of epithelium-specific cadherin (red) and biomarkers of EMT as vimentin (green) expression F) Ward’s clustering cytopathologic features in CS and CP. G) mutational profile in order of Kras mutation comparing liquid and tumour biopsy and CTCs with ct-DNA in same colon cancer patients. H) workflow starts from blood collection through short time culture to ML application. I) Neural model incorporates a linear, non-linear, normalization layer linked together. L) Decision boundary or confusion region between the two groups in order of S-phase input M) Performance of the neural network model N) 3D representation of the model output as a function of cancer features as S-phase, atypical cells (Atc) and cells organized in clusters (CCF)