The coming era of nudge drugs for cancer

We propose an emerging strategy for advanced cancer treatment based on progressive, stepwise remodeling of tumor microenvironments (TMEs). TMEs are variable but show conserved archetypes across patients and tissue origins. Deep learning over single-cell atlases collected from perturbed tumors can uncover gene and cellular networks shifting between archetypes. This allows for designing “nudge” or “state-shifting” drugs whose sequential application achieves stepwise transformation of a TME from an adverse to a more favorable state, dismantling deleterious tumor-host interactions to achieve patient remission.