Fig. 5: In vitro experimental validation of Morpheus predictions.

a, T-cell transwell migration assay for assessing the effect of Morpheus-derived perturbation strategies on CD8⁺ T-cell infiltration into an in vitro tumour compartment. Human PBMCs are placed into the top chamber, and a human tumour cell line (A375 for melanoma and HCT116 for CRC) is placed into the bottom chamber. We measure CD8⁺ T-cell infiltration into the bottom chamber after 4 h in the presence or absence of signalling perturbations predicted by Morpheus. Signalling perturbations include both signalling protein addition and blocking antibodies, which are indicated by α/anti. b, Log fold change in CD8⁺ T-cell abundance within the lower chamber containing A375 melanoma cells relative to CD8⁺ T-cell abundance in unperturbed controls. CXCL9 and {CXCL10, αCCL18, αCCL22} are Morpheus predicted infiltration strategies, while the CXCL10 addition alone strategy is shown for comparison to CXCL9 alone. c, Log fold change in CD8⁺ T-cell abundance within the lower chamber containing HCT116 CRC cells relative to CD8⁺ T-cell abundance in unperturbed control. {αPD-1, αPD-L1, αCXCR4} and {αPD-1, αPD-L1, αCXCR4, αCYR61} are Morpheus predicted strategies. The αPD-1 and αPD-L1 strategies, blocking PD-1 or PD-L1, are clinical immunotherapy strategies shown for comparison. Each perturbation trial was normalized to its paired control trial. *P < 5 × 10⁻² and **P < 1 × 10⁻². Two-sided paired t-tests used to assess significance; see Supplementary Tables 6 and 7 for raw data. The error bars represent the mean ± s.e.m. of independent biological replicates and dots indicate individual n = 8 and n = 9 replicate values for b and c, respectively (n = 12 for bars 2–3 in c).