Tej Pandya, Maria Zagorulya, Michelle Leung, Marcellus Augustine, Lydia Liu, Aino‐Maija Leppä, Ulysse Baruchel, Sin Wi Ng, Tamara Klockner, Miriam Mugabo, Anthony Griffen, Oleg Blyuss, Chrysante S. Iliakis, Amalie Grenov, Kerstin Haase, David C. Muller, K Chan, Jincheng Wu, Vernon A. Burk, Neil Wright, Alix Le Marois, Ekaterina Pazukhina, Sophia Ward, Hubert Slawinski, Marc Pelletier, C L Murphy, Matthew D. Park, Thomas Snoeks, Alejandro Suarez-Bonnet, Simon L. Priestnall, Alexandros Hardas, Charlotte Grieco, Ami Archer, Alpkaan Celik, Alejandro Jiménez-Sánchez, Rachel Scott, Hana Zahed, Léa Montégut, Rafael Meza, Clinton H. Durney, Stephen Lam, Takahiro Karasaki, Roel C.H. Vermeulen, Huilei Xu, Pablo Serrano‐Fernández, Tatjana Crnogorac-Jurcevic, Usha Menon, Sophia Apostolidou, Alexey Zaikin, Richard Gunu, Harry J. Whitwell, Zhe Huang, Zonglun Li, Xin Hu, Bo Zhu, Liming Li, MD Chirlaque, Marcela Guevara, P Martijn Kolijn, Aghiles Guenoun, Neeloffer Mookherjee, Mattias Johansson, Ziqiao Wang, Nilanjan Chatterjee, Chao‐Hua Chiu, Z M Chen, Dana Pe’er, Erik Sahai, Saskia Freytag, Andreas Wack, Marc J. Gunter, Miriam Merad, J Zhang, C. Carlsten, Pan-Chyr Yang, H Chen, Elizabeth A. Platz, Lindsay M. LaFave, Karl Smith-Byrne, Mariam Jamal-Hanjani, Kevin Litchfield, Nuno R. Nene, Nicholas McGranahan, Eva Grönroos, W. Hill, Clare E. Weeden, Charles Swanton
Predicting lung cancer risk would enhance prevention trials. Although the Canakinumab Anti-inflammatory Thrombosis Outcome Study (CANTOS) trial demonstrated reduced lung cancer incidence with interleukin (IL)-1β inhibition, the high number needed to treat (NNT) to prevent lung cancer limits its use in unselected populations. Using machine learning, we identified a 14-protein plasma signature predicting lung cancer more than 5 years before diagnosis. The signature, validated across eight cohorts, was elevated in current smokers and individuals exposed to particulate matter (PM) and linked to lung myeloid and alveolar cells. In epidermal growth factor receptor ( EGFR )-driven lung adenocarcinoma, diverse epithelial lineages converged on a keratin8 + /claudin4 + alveolar transitional state (KAC), whose transcriptional programs correlated with signature emergence. Components of the signature were induced by PM, oncogenic EGFR , or IL-1β, whereas IL-1β inhibition restrained PM-driven KAC expansion and early tumorigenesis. In CANTOS, the signature identified individuals who seemed to benefit more from anti-IL-1β therapy, lowering the NNT threshold and nominating circulating signals of tumor promotion for prevention.