Find this paper

Abstract

Treatment toxicity constrains the use of effective cancer therapies, but its clinical and genetic determinants remain poorly defined. We developed a large language model-based approach to produce MSK-Tox, a pan-cancer resource capturing the incidence, temporality, and grade of toxicity to anti-cancer therapy across more than 50,000 patients. Analysis of six representative adverse events--pneumonitis, adrenal insufficiency, liver toxicity, colitis, hyperthyroidism and hypothyroidism--revealed distinct toxicity landscapes shaped by cancer type, treatment regimen, and clinical context. Pretreatment clinical features enabled individualized prediction of toxicity risk across adverse events, supporting risk assessment before therapy initiation. Beyond clinical predictors, we identified two modes of germline susceptibility to treatment toxicity: an organintrinsic mode, in which germline variation confers risk across systemic therapies, exemplified by a regulatory variant near FOXE1 associated with hypothyroidism; and an immune-mediated mode, confined to immune checkpoint inhibitor-treated patients, in which HLA-DRB1*15 was a major determinant of adrenal insufficiency. Notably, the same allele predisposes to multiple sclerosis in individuals without cancer, indicating that immune checkpoint inhibition unmasks a latent autoimmune predisposition. These findings provide an empirical basis for a new precision safety paradigm for predicting who will be harmed by a therapy on the same principles that guide prediction of therapeutic benefit.

Posts about this paper

No posts about this paper yet. Mention it with @ in a post to start the conversation.